{ "name": "stack.tools production AI stacks dataset", "url": "https://stack.tools/data/stacks.json", "license": "CC BY 4.0 with attribution to stack.tools", "generated_at": "2026-08-22", "company_count": 100, "fact_count": 222, "companies": [ { "slug": "1up", "name": "1up", "website": "https://1up.ai", "what": "AI answer engine automating RFPs and security questionnaires for sales teams.", "profile": "1up improved its Knowledge Automation System by choosing Pinecone as its vector database, streamlining data storage and retrieval for RAG.", "facts": [ { "tool": "Pinecone", "vendor": "Pinecone", "layer": "rag-vector", "usage": "Streamlines sales knowledge storage and retrieval for RAG.", "status": "current", "confidence": "high", "last_verified": "2026-07-27", "added": "2026-07-27", "evidence": [ { "url": "https://www.pinecone.io/customers/1up/", "type": "case-study", "date": "2026-07", "quote": "By choosing Pinecone as its vector database, 1up improved its Knowledge Automation System, streamlining data storage and retrieval." } ] } ] }, { "slug": "allspice", "name": "Allspice", "website": "https://www.allspicelabs.com", "what": "AI cooking tools for recipe publishers.", "profile": "Allspice normalizes ingredient and recipe data and powers its chatbot with semantic matching, with Pinecone handling vector storage and similarity search.", "facts": [ { "tool": "Pinecone", "vendor": "Pinecone", "layer": "rag-vector", "usage": "Handles vector storage and similarity search for recipe data.", "status": "current", "confidence": "high", "last_verified": "2026-07-27", "added": "2026-07-27", "evidence": [ { "url": "https://www.pinecone.io/customers/allspice/", "type": "case-study", "date": "2026-07", "quote": "Pinecone handles vector storage and similarity search." } ] } ] }, { "slug": "amgen", "name": "Amgen", "website": "https://www.amgen.com", "what": "Global biotechnology company developing therapeutics.", "profile": "Amgen applies GPT-5 across its enterprise workflows.", "facts": [ { "tool": "GPT-5", "vendor": "OpenAI", "layer": "models", "usage": "Applied across enterprise workflows.", "status": "current", "confidence": "medium", "last_verified": "2026-07-27", "added": "2026-07-27", "evidence": [ { "url": "https://openai.com/index/gpt-5-amgen/", "type": "case-study", "date": "2025-08-07", "quote": "How Amgen uses GPT-5" } ] } ] }, { "slug": "applied-compute", "name": "Applied Compute", "website": "https://appliedcompute.com", "what": "Startup building reinforcement-learning-trained enterprise AI models.", "profile": "Applied Compute runs reinforcement learning rollouts, grading, and inference on Modal, using Modal Functions for inexpensive serverless fan-out without a dedicated cluster.", "facts": [ { "tool": "Modal", "vendor": "Modal", "layer": "training", "usage": "Executes reinforcement learning rollouts, grading, and inference workloads.", "status": "current", "confidence": "high", "last_verified": "2026-07-27", "added": "2026-07-27", "evidence": [ { "url": "https://modal.com/blog/applied-compute-reinforcement-learning", "type": "case-study", "date": "2026-05-20", "quote": "Applied Compute makes use of Modal Functions to provide inexpensive serverless fan-out without requiring a dedicated cluster." } ] } ] }, { "slug": "aquant", "name": "Aquant", "website": "https://www.aquant.ai", "what": "Agentic AI platform for servicing complex equipment.", "profile": "Aquant anchors its agentic toolchain on Pinecone, which powers semantic search and retrieval across its service intelligence platform.", "facts": [ { "tool": "Pinecone", "vendor": "Pinecone", "layer": "rag-vector", "usage": "Powers semantic search and retrieval in its agentic platform.", "status": "current", "confidence": "high", "last_verified": "2026-07-27", "added": "2026-07-27", "evidence": [ { "url": "https://www.pinecone.io/customers/aquant/", "type": "case-study", "date": "2026-07", "quote": "Pinecone is now an anchor of Aquant" } ] } ] }, { "slug": "avalara", "name": "Avalara", "website": "https://www.avalara.com", "what": "Tax compliance automation software company.", "profile": "Avalara's product teams translate plain language into working prototypes with v0 for design and validation of AI-native customer workflows.", "facts": [ { "tool": "v0", "vendor": "Vercel", "layer": "ai-coding", "usage": "Translates plain language into working prototypes for validation.", "status": "current", "confidence": "high", "last_verified": "2026-07-27", "added": "2026-07-27", "evidence": [ { "url": "https://vercel.com/customers/How-avalara-turns-pipedreams-into-patent-pending-with-v0", "type": "case-study", "date": "2026-02-28", "quote": "Enter Vercel’s v0 , which translates plain language into working prototypes." } ] } ] }, { "slug": "box", "name": "Box", "website": "https://www.box.com", "what": "Cloud content management platform for enterprises.", "profile": "Box built its eval practice in Braintrust, curating datasets, running programmatic evaluations, and comparing model performance for its AI agent.", "facts": [ { "tool": "Braintrust", "vendor": "Braintrust", "layer": "evals-observability", "usage": "Runs programmatic evaluations and dataset curation for its AI agent.", "status": "current", "confidence": "high", "last_verified": "2026-07-27", "added": "2026-07-27", "evidence": [ { "url": "https://www.braintrust.dev/customers/box", "type": "case-study", "date": "2026-07", "quote": "Their team built an eval practice in Braintrust that helps them curate datasets" } ] } ] }, { "slug": "browserbase", "name": "Browserbase", "website": "https://browserbase.com", "what": "Headless browser infrastructure for AI agents.", "profile": "Browserbase uses Braintrust to run benchmarks and observe browser-agent model performance — the metrics its customers care about.", "facts": [ { "tool": "Braintrust", "vendor": "Braintrust", "layer": "evals-observability", "usage": "Runs benchmarks and evaluates browser-agent model performance.", "status": "current", "confidence": "high", "last_verified": "2026-07-27", "added": "2026-07-27", "evidence": [ { "url": "https://www.braintrust.dev/customers/browserbase", "type": "case-study", "date": "2026-07", "quote": "what Browserbase uses Braintrust to observe and understand" } ] } ] }, { "slug": "c-h-robinson", "name": "C.H. Robinson", "website": "https://www.chrobinson.com", "what": "Global logistics and freight brokerage company.", "profile": "C.H. Robinson's GenAI engineering team builds its AI agents on the open-source LangChain framework for interoperability, prototypes and debugs them with LangGraph Studio, and uses LangSmith as the first line of defense in testing.", "facts": [ { "tool": "LangChain", "vendor": "LangChain", "layer": "orchestration", "usage": "Builds AI agents for interoperability and model switching.", "status": "current", "confidence": "high", "last_verified": "2026-07-27", "added": "2026-07-27", "evidence": [ { "url": "https://blog.langchain.dev/customers-chrobinson/", "type": "case-study", "date": "2025-03-10", "quote": "C.H. Robinson’s GenAI engineering team started by building their AI agents using langchain (the open-source framework) for maximum interoperability." } ] }, { "tool": "LangGraph", "vendor": "LangChain", "layer": "orchestration", "usage": "Orchestrates shipment agents prototyped and debugged in LangGraph Studio.", "status": "current", "confidence": "high", "last_verified": "2026-07-27", "added": "2026-07-27", "evidence": [ { "url": "https://blog.langchain.dev/customers-chrobinson/", "type": "case-study", "date": "2025-03-10", "quote": "The visual LangGraph Studio also helped their engineers prototype and debug complex agent interactions, saving them development time." } ] }, { "tool": "LangSmith", "vendor": "LangChain", "layer": "evals-observability", "usage": "Provides real-time observability and error tracking during testing.", "status": "current", "confidence": "high", "last_verified": "2026-07-27", "added": "2026-07-27", "evidence": [ { "url": "https://blog.langchain.dev/customers-chrobinson/", "type": "case-study", "date": "2025-03-10", "quote": "LangSmith was their first line of defense in the testing process" } ] } ] }, { "slug": "cartesia", "name": "Cartesia", "website": "https://cartesia.ai", "what": "Real-time voice AI company building state-space models.", "profile": "Cartesia serves millions of audio minutes daily on Together AI with a custom SSM inference engine, and trains its voice models on Together's multi-node GPU clusters scheduled via Slurm.", "facts": [ { "tool": "Together AI", "vendor": "Together AI", "layer": "inference", "usage": "Runs a custom SSM inference engine for real-time voice.", "status": "current", "confidence": "high", "last_verified": "2026-07-27", "added": "2026-07-27", "evidence": [ { "url": "https://www.together.ai/customers/cartesia", "type": "case-study", "date": "2026-07", "quote": "To support these requirements — and serve millions of audio minutes daily — Cartesia uses Together AI." } ] }, { "tool": "Together GPU Clusters", "vendor": "Together AI", "layer": "training", "usage": "Trains voice models on multi-node clusters scheduled via Slurm.", "status": "current", "confidence": "high", "last_verified": "2026-07-27", "added": "2026-07-27", "evidence": [ { "url": "https://www.together.ai/customers/cartesia", "type": "case-study", "date": "2026-07", "quote": "Cartesia also uses Together’s H100 80GB clusters for model training, with multi-node workloads scheduled via Slurm." } ] } ] }, { "slug": "caylent", "name": "Caylent", "website": "https://caylent.com", "what": "AWS premier consulting partner delivering cloud and AI services.", "profile": "Caylent has deployed Claude Enterprise company-wide, made Claude Code its engineers' primary development tooling, and rebuilt Accelerate, its agentic delivery system for migrations, on the Claude Agent SDK.", "facts": [ { "tool": "Claude Code", "vendor": "Anthropic", "layer": "ai-coding", "usage": "Serves as engineers' primary development tooling.", "status": "current", "confidence": "high", "last_verified": "2026-07-27", "added": "2026-07-27", "evidence": [ { "url": "https://claude.com/customers/caylent", "type": "case-study", "date": "2026-07", "quote": "made Claude Code our engineers’ primary development tooling." } ] }, { "tool": "Claude Enterprise", "vendor": "Anthropic", "layer": "internal-ai", "usage": "Deployed company-wide for daily operational use.", "status": "current", "confidence": "high", "last_verified": "2026-07-27", "added": "2026-07-27", "evidence": [ { "url": "https://claude.com/customers/caylent", "type": "case-study", "date": "2026-07", "quote": "We’ve deployed Claude Enterprise company-wide" } ] }, { "tool": "Claude Agent SDK", "vendor": "Anthropic", "layer": "orchestration", "usage": "Powers the Accelerate agentic delivery system for migrations.", "status": "current", "confidence": "high", "last_verified": "2026-07-27", "added": "2026-07-27", "evidence": [ { "url": "https://claude.com/customers/caylent", "type": "case-study", "date": "2026-07", "quote": "The team rebuilt Accelerate on the Claude Agent SDK," } ] } ] }, { "slug": "chai-discovery", "name": "Chai Discovery", "website": "https://www.chaidiscovery.com", "what": "AI company for molecular structure prediction and drug discovery.", "profile": "Chai Discovery runs computational biology workflows from research through production on Modal, with Modal Volumes providing a shared high-performance distributed file system for datasets.", "facts": [ { "tool": "Modal", "vendor": "Modal", "layer": "inference", "usage": "Runs computational biology workflows from research through production.", "status": "current", "confidence": "high", "last_verified": "2026-07-27", "added": "2026-07-27", "evidence": [ { "url": "https://modal.com/blog/seamless-computational-bio-at-chai-discovery", "type": "case-study", "date": "2026-01-15", "quote": "With Modal, building infrastructure has shifted from being an imperative task to a declarative one." }, { "url": "https://modal.com/blog/seamless-computational-bio-at-chai-discovery", "type": "case-study", "date": "2026-01-15", "quote": "With Modal Volumes it’s downloaded once, instantly available everywhere, and scales to thousands of queries." } ] } ] }, { "slug": "clay", "name": "Clay", "website": "https://www.clay.com", "what": "Go-to-market data enrichment and outbound automation platform.", "profile": "Clay uses OpenAI models to classify and extract attributes from company websites and account data.", "facts": [ { "tool": "OpenAI models", "vendor": "OpenAI", "layer": "models", "usage": "Classifies and extracts attributes from websites and account data.", "status": "current", "confidence": "high", "last_verified": "2026-07-27", "added": "2026-07-27", "evidence": [ { "url": "https://openai.com/index/frontier-builders/", "type": "case-study", "date": "2026-07", "quote": "OpenAI models are used to classify and extract attributes from company websites" } ] } ] }, { "slug": "cloudflare", "name": "Cloudflare", "website": "https://www.cloudflare.com", "what": "Internet infrastructure and developer platform.", "profile": "Cloudflare's dashboard agent — which now invokes four MCP suites to act on a user's behalf — is measured in Braintrust with an LLM-as-a-judge scoring whole conversations for resolution, broken down by product area. The eval diff runs only when a skill, prompt, or recipe changes and blocks regressions in CI/CD; the same suite benchmarks sub-agents against models of varying capacity, and covers Cloudflare's internal runbook automation and AI code reviewer.", "facts": [ { "tool": "Braintrust", "vendor": "Braintrust", "layer": "evals-observability", "usage": "Evaluates the dashboard agent and gates skill changes in CI/CD.", "status": "current", "confidence": "high", "last_verified": "2026-08-22", "added": "2026-08-22", "evidence": [ { "url": "https://www.braintrust.dev/customers/cloudflare", "type": "case-study", "date": "2026-08", "quote": "That was something we leaned on Braintrust for." } ] } ] }, { "slug": "code-and-theory", "name": "Code and Theory", "website": "https://www.codeandtheory.com", "what": "Digital-first creative and technology agency.", "profile": "Code and Theory replaced wireframes with prompt-to-code workflows in v0, cutting its time to prototype.", "facts": [ { "tool": "v0", "vendor": "Vercel", "layer": "ai-coding", "usage": "Replaces wireframes with prompt-to-code workflows.", "status": "current", "confidence": "high", "last_verified": "2026-07-27", "added": "2026-07-27", "evidence": [ { "url": "https://vercel.com/customers/how-code-and-theory-cut-time-to-prototype-75-with-v0", "type": "case-study", "date": "2026-06-09", "quote": "prompt-to-code workflows in v0" } ] } ] }, { "slug": "coursera", "name": "Coursera", "website": "https://www.coursera.org", "what": "Online learning platform offering courses and degrees.", "profile": "Coursera evaluates its AI features and maintains continuous quality awareness of production performance through Braintrust.", "facts": [ { "tool": "Braintrust", "vendor": "Braintrust", "layer": "evals-observability", "usage": "Evaluates AI features and monitors production quality.", "status": "current", "confidence": "high", "last_verified": "2026-07-27", "added": "2026-07-27", "evidence": [ { "url": "https://www.braintrust.dev/customers/coursera", "type": "case-study", "date": "2026-07", "quote": "With evaluation infrastructure in place through Braintrust, Coursera maintains continuous quality awareness" } ] } ] }, { "slug": "cresta", "name": "Cresta", "website": "https://cresta.com", "what": "AI platform for contact center agent assistance and insights.", "profile": "Cresta serves real-time LLM applications on Fireworks AI, valuing its low-latency, high-throughput model serving.", "facts": [ { "tool": "Fireworks AI", "vendor": "Fireworks AI", "layer": "inference", "usage": "Serves low-latency LLMs for real-time contact center applications.", "status": "current", "confidence": "high", "last_verified": "2026-07-27", "added": "2026-07-27", "evidence": [ { "url": "https://fireworks.ai/blog/story-cresta-knowledge-assist", "type": "case-study", "date": "2024-12-08", "quote": "The low-latency, high-throughput serving of LLMs has been particularly valuable, as latency is crucial for our real-time applications." } ] } ] }, { "slug": "cursor", "name": "Cursor (Anysphere)", "website": "https://cursor.com", "what": "AI code editor and agentic coding platform (Agent, Tab, Composer, Bugbot, Cloud Agents).", "profile": "Cursor, built by Anysphere, is the most prominent AI coding company and one of the fastest-growing software businesses ever. Its stack is a deliberate hybrid: it routes to every frontier lab (Anthropic, OpenAI, Google, xAI) while training its own frontier models (Composer, Tab) with large-scale RL on thousands of NVIDIA Blackwell GPUs, on in-house infrastructure (Anyrun, custom PyTorch+Ray pipelines). Build-vs-buy philosophy: build the model/RL/eval layer, buy best-of-breed infra (Turbopuffer, Fireworks, AWS, Datadog).", "facts": [ { "tool": "Composer", "vendor": "Anysphere", "layer": "models", "usage": "Flagship in-house model family specialized for agentic coding.", "status": "current", "confidence": "high", "last_verified": "2026-07-26", "added": "2026-07-26", "evidence": [ { "url": "https://cursor.com/blog/composer", "type": "eng-blog", "date": "2025-10-29", "quote": "It is specialized for software engineering through reinforcement learning (RL) in a diverse range of development environments." }, { "url": "https://cursor.com/docs/models", "type": "docs", "date": "2026-07", "quote": "Composer 2.5 is Cursor's own model, trained to be highly capable for agentic coding." } ] }, { "tool": "Claude", "vendor": "Anthropic", "layer": "models", "usage": "Offered as a first-class frontier model choice in Agent and chat.", "status": "current", "confidence": "high", "last_verified": "2026-07-26", "added": "2026-07-26", "evidence": [ { "url": "https://cursor.com/docs/models", "type": "docs", "date": "2026-07", "quote": "Claude Sonnet 5" }, { "url": "https://simonwillison.net/2025/May/11/cursor-security/", "type": "press", "date": "2025-05-11", "quote": "make API calls out to OpenAI, Anthropic, Gemini and xAI" } ] }, { "tool": "GPT-5 series", "vendor": "OpenAI", "layer": "models", "usage": "OpenAI frontier and Codex models offered in the editor's model picker.", "status": "current", "confidence": "high", "last_verified": "2026-07-27", "added": "2026-07-26", "evidence": [ { "url": "https://cursor.com/docs/models", "type": "docs", "date": "2026-07", "quote": "Cursor supports frontier models from OpenAI, Anthropic, Google, SpaceXAI, and more" }, { "url": "https://openai.com/index/gpt-5-cursor/", "type": "case-study", "date": "2025-08-07", "quote": "How Cursor uses GPT-5" }, { "url": "https://www.latent.space/p/cursor", "type": "podcast", "date": "2023-08-22", "quote": "GPT-3.5 and GPT-4 to ask questions and receive answers based on their code" } ] }, { "tool": "Grok 4.5", "vendor": "SpaceXAI (xAI)", "layer": "models", "usage": "Jointly trained frontier model offered as a first-party option across plans.", "status": "current", "confidence": "high", "last_verified": "2026-07-26", "added": "2026-07-26", "evidence": [ { "url": "https://cursor.com/docs/models", "type": "docs", "date": "2026-07", "quote": "Grok 4.5 is jointly trained by Cursor and SpaceXAI for long-running coding and knowledge work." } ] }, { "tool": "Kimi K2.5", "vendor": "Moonshot AI", "layer": "models", "usage": "Open-weights base model that Composer's training starts from.", "status": "current", "confidence": "high", "last_verified": "2026-07-26", "added": "2026-07-26", "evidence": [ { "url": "https://cursor.com/blog/composer-2-technical-report", "type": "eng-blog", "date": "2026-03-27", "quote": "Continued pretraining on an open base model, Kimi K2.5" } ] }, { "tool": "Fireworks AI", "vendor": "Fireworks AI", "layer": "inference", "usage": "Inference host serving Cursor's custom fine-tuned models.", "status": "current", "confidence": "high", "last_verified": "2026-07-27", "added": "2026-07-26", "evidence": [ { "url": "https://fireworks.ai/blog/cursor", "type": "case-study", "date": "2024-06-23", "quote": "Fireworks deployed Cursor's special fine-tune of Llama-3-70b for the coding task 'Fast Apply' using the speculative API flag." }, { "url": "https://fireworks.ai/blog/Cursor-Composer-2", "type": "case-study", "date": "2026-06-26", "quote": "Fireworks provides the inference layer that makes these RL loops practical." }, { "url": "https://simonwillison.net/2025/May/11/cursor-security/", "type": "press", "date": "2025-05-11", "quote": "host their own custom models on Fireworks" }, { "url": "https://cursor.com/blog/composer-2-technical-report", "type": "eng-blog", "date": "2026-03-27", "quote": "We'd also like to thank Fireworks and Colfax for their collaboration and partnership." } ] }, { "tool": "AWS", "vendor": "Amazon Web Services", "layer": "inference", "usage": "Primary cloud for the backend that routes AI traffic to model providers.", "status": "current", "confidence": "high", "last_verified": "2026-07-26", "added": "2026-07-26", "evidence": [ { "url": "https://newsletter.pragmaticengineer.com/p/cursor", "type": "press", "date": "2025-06-10", "quote": "We are very much a 'cloud shop.' We mostly rely on AWS and then Azure for inference." }, { "url": "https://simonwillison.net/2025/May/11/cursor-security/", "type": "press", "date": "2025-05-11", "quote": "AWS for primary infrastructure, Azure and GCP for \"some secondary infrastructure\"" } ] }, { "tool": "Together GPU Clusters", "vendor": "Together AI", "layer": "inference", "usage": "Runs production inference on dedicated NVIDIA Blackwell GPU clusters.", "status": "current", "confidence": "high", "last_verified": "2026-07-27", "added": "2026-07-27", "evidence": [ { "url": "https://www.together.ai/customers/cursor", "type": "case-study", "date": "2026-07", "quote": "Cursor partnered with Together AI to deploy production inference on NVIDIA Blackwell" } ] }, { "tool": "Model Context Protocol", "vendor": "Anthropic (open standard)", "layer": "orchestration", "usage": "Extensibility layer connecting the agent to external tools and data.", "status": "current", "confidence": "high", "last_verified": "2026-07-26", "added": "2026-07-26", "evidence": [ { "url": "https://cursor.com/docs/mcp", "type": "docs", "date": "2026-07", "quote": "Model Context Protocol (MCP) enables Cursor to connect to external tools and data sources." } ] }, { "tool": "turbopuffer", "vendor": "turbopuffer", "layer": "rag-vector", "usage": "Hosted vector store for codebase-index embeddings powering semantic code retrieval.", "status": "current", "confidence": "high", "last_verified": "2026-07-26", "added": "2026-07-26", "evidence": [ { "url": "https://turbopuffer.com/customers/cursor", "type": "case-study", "date": "2026-01", "quote": "turbopuffer is one of the few pieces of infrastructure we haven't had to worry about as we've scaled." }, { "url": "https://simonwillison.net/2025/May/11/cursor-security/", "type": "press", "date": "2025-05-11", "quote": "At our server, we chunk and embed the files, and store the embeddings in Turbopuffer" } ] }, { "tool": "Custom codebase indexing (Merkle trees + embeddings)", "vendor": "Anysphere", "layer": "rag-vector", "usage": "Indexes codebases with client-side change detection and server-side embeddings for search.", "status": "current", "practice": true, "confidence": "high", "last_verified": "2026-07-26", "added": "2026-07-26", "evidence": [ { "url": "https://cursor.com/blog/secure-codebase-indexing", "type": "eng-blog", "date": "2026-01-27", "quote": "features a cryptographic hash of every file, along with hashes of each folder" }, { "url": "https://cursor.com/blog/secure-codebase-indexing", "type": "eng-blog", "date": "2026-01-27", "quote": "it improved response accuracy by 12.5% on average" } ] }, { "tool": "Cursor Bench / BugBench (internal evals)", "vendor": "Anysphere", "layer": "evals-observability", "usage": "Internal benchmark suites built from real usage gate model releases.", "status": "current", "practice": true, "confidence": "high", "last_verified": "2026-07-26", "added": "2026-07-26", "evidence": [ { "url": "https://cursor.com/blog/composer", "type": "eng-blog", "date": "2025-10-29", "quote": "Our benchmark, Cursor Bench, consists of real agent requests from engineers and researchers at Cursor" }, { "url": "https://cursor.com/blog/building-bugbot", "type": "eng-blog", "date": "2026-01-15", "quote": "offline using BugBench, a curated benchmark of real code diffs" } ] }, { "tool": "Datadog", "vendor": "Datadog", "layer": "evals-observability", "usage": "Primary monitoring and observability platform across serving infrastructure.", "status": "current", "confidence": "medium", "last_verified": "2026-07-26", "added": "2026-07-26", "evidence": [ { "url": "https://newsletter.pragmaticengineer.com/p/cursor", "type": "press", "date": "2025-06-10", "quote": "heavy users and find the developer experience of Datadog vastly superior to the alternatives" } ] }, { "tool": "Ray", "vendor": "Anyscale (open source)", "layer": "training", "usage": "Underpins the custom asynchronous reinforcement-learning training infrastructure with PyTorch.", "status": "current", "confidence": "high", "last_verified": "2026-07-26", "added": "2026-07-26", "evidence": [ { "url": "https://cursor.com/blog/composer", "type": "eng-blog", "date": "2025-10-29", "quote": "We built custom training infrastructure leveraging PyTorch and Ray to power asynchronous reinforcement learning at scale." }, { "url": "https://cursor.com/careers/software-engineer-ml-research", "type": "job-post", "date": "2026-07", "quote": "Build our distributed training, inference, and RL infrastructure" } ] }, { "tool": "NVIDIA GPUs (Blackwell)", "vendor": "NVIDIA", "layer": "training", "usage": "Trains mixture-of-experts models at low precision on large GPU clusters.", "status": "current", "confidence": "high", "last_verified": "2026-07-26", "added": "2026-07-26", "evidence": [ { "url": "https://cursor.com/blog/composer", "type": "eng-blog", "date": "2025-10-29", "quote": "allowing us to scale training to thousands of NVIDIA GPUs with minimal communication cost" }, { "url": "https://cursor.com/blog/composer-2-technical-report", "type": "eng-blog", "date": "2026-03-27", "quote": "Custom low-precision kernels for efficient MoE training on Blackwell GPUs" } ] }, { "tool": "Anyrun", "vendor": "Anysphere", "layer": "training", "usage": "Internal compute platform running sandboxed cloud environments for RL rollouts.", "status": "current", "confidence": "high", "last_verified": "2026-07-26", "added": "2026-07-26", "evidence": [ { "url": "https://cursor.com/blog/composer-2-technical-report", "type": "eng-blog", "date": "2026-03-27", "quote": "Anyrun, our internal compute platform for running hundreds of thousands of sandboxed coding environments" }, { "url": "https://cursor.com/blog/composer", "type": "eng-blog", "date": "2025-10-29", "quote": "running hundreds of thousands of concurrent sandboxed coding environments in the cloud" } ] }, { "tool": "Online RL pipeline (Tab model)", "vendor": "Anysphere", "layer": "training", "usage": "Trains the Tab model continuously on live accept and reject feedback.", "status": "current", "practice": true, "confidence": "high", "last_verified": "2026-07-26", "added": "2026-07-26", "evidence": [ { "url": "https://cursor.com/blog/tab-rl", "type": "eng-blog", "date": "2025-09-12", "quote": "rolling out new models to users frequently throughout the day and using that data for training" }, { "url": "https://cursor.com/careers/software-engineer-ml-research", "type": "job-post", "date": "2026-07", "quote": "train frontier coding agents and scale RL on real user data" } ] }, { "tool": "In-house expert data curation", "vendor": "Anysphere", "layer": "data-labeling", "usage": "Engineers hand-curate gold-standard solutions for evaluation and benchmark data.", "status": "current", "practice": true, "confidence": "high", "last_verified": "2026-07-26", "added": "2026-07-26", "evidence": [ { "url": "https://cursor.com/blog/composer", "type": "eng-blog", "date": "2025-10-29", "quote": "along with hand-curated optimal solutions to these requests" } ] }, { "tool": "Sandboxed terminals (Seatbelt / seccomp + Landlock)", "vendor": "Anysphere", "layer": "guardrails", "usage": "Runs agent shell commands in OS-level sandboxes with restricted filesystem access.", "status": "current", "practice": true, "confidence": "high", "last_verified": "2026-07-26", "added": "2026-07-26", "evidence": [ { "url": "https://cursor.com/blog/agent-sandboxing", "type": "eng-blog", "date": "2026-02-18", "quote": "Seccomp blocks unsafe syscalls, while Landlock enforces filesystem restrictions." }, { "url": "https://cursor.com/blog/agent-sandboxing", "type": "eng-blog", "date": "2026-02-18", "quote": "A mistaken agent can delete databases, ship broken code, or leak secrets." } ] }, { "tool": "Cursor", "vendor": "Anysphere", "layer": "ai-coding", "usage": "Builds Cursor with Cursor through company-wide dogfooding.", "status": "current", "confidence": "high", "last_verified": "2026-07-26", "added": "2026-07-26", "evidence": [ { "url": "https://cursor.com/blog/composer", "type": "eng-blog", "date": "2025-10-29", "quote": "real agent requests from engineers and researchers at Cursor" }, { "url": "https://newsletter.pragmaticengineer.com/p/cursor", "type": "press", "date": "2025-06-10", "quote": "whether they wrote it by hand, or had Cursor generate it" } ] }, { "tool": "Bugbot", "vendor": "Anysphere", "layer": "ai-coding", "usage": "Reviews every internal pull request with Cursor's own AI reviewer.", "status": "current", "confidence": "high", "last_verified": "2026-07-26", "added": "2026-07-26", "evidence": [ { "url": "https://cursor.com/blog/building-bugbot", "type": "eng-blog", "date": "2026-01-15", "quote": "We also run Bugbot on all internal code at Cursor." } ] } ] }, { "slug": "customgpt-ai", "name": "CustomGPT.ai", "website": "https://customgpt.ai", "what": "No-code platform for building RAG agents on business content.", "profile": "CustomGPT.ai relies on Pinecone as foundational infrastructure for accurate, production-grade vector retrieval behind its real-time RAG and agent workloads.", "facts": [ { "tool": "Pinecone", "vendor": "Pinecone", "layer": "rag-vector", "usage": "Provides production-grade vector retrieval for real-time RAG agents.", "status": "current", "confidence": "high", "last_verified": "2026-07-27", "added": "2026-07-27", "evidence": [ { "url": "https://www.pinecone.io/customers/customgpt-ai/", "type": "case-study", "date": "2026-07", "quote": "We trust Pinecone to provide the foundational infrastructure we rely on for accurate, production-grade vector retrieval at scale." } ] } ] }, { "slug": "cyera", "name": "Cyera", "website": "https://www.cyera.io", "what": "Data security platform that discovers, classifies, and maps access to enterprise data.", "profile": "Cyera standardized on Claude Cowork as the front door to its stack — roughly forty connected tools plus four or five homegrown MCP servers — so non-technical teams can run self-service analytics over a governed Snowflake semantic layer and hand ticket triage to their own agents. R&D stays on Claude Code, and company-wide access runs on Claude Enterprise.", "facts": [ { "tool": "Claude Cowork", "vendor": "Anthropic", "layer": "internal-ai", "usage": "Front door to connected business tools for non-engineering workflows.", "status": "current", "confidence": "high", "last_verified": "2026-08-22", "added": "2026-08-22", "evidence": [ { "url": "https://claude.com/customers/cyera-qa", "type": "case-study", "date": "2026-08", "quote": "everyone else using Claude Cowork, which Cyera has connected to 40 tools across its stack." } ] }, { "tool": "Claude Enterprise", "vendor": "Anthropic", "layer": "internal-ai", "usage": "Company-wide Claude access across engineering and business teams.", "status": "current", "confidence": "high", "last_verified": "2026-08-22", "added": "2026-08-22", "evidence": [ { "url": "https://claude.com/customers/cyera-qa", "type": "case-study", "date": "2026-08", "quote": "scales agentic AI across 1,500 employees with Claude Enterprise." } ] }, { "tool": "Claude Code", "vendor": "Anthropic", "layer": "ai-coding", "usage": "Primary agentic coding tool for the R&D organization.", "status": "current", "confidence": "high", "last_verified": "2026-08-22", "added": "2026-08-22", "evidence": [ { "url": "https://claude.com/customers/cyera-qa", "type": "case-study", "date": "2026-08", "quote": "R&D using Claude Code primarily" } ] } ] }, { "slug": "decagon", "name": "Decagon", "website": "https://decagon.ai", "what": "AI customer service agents for enterprises.", "profile": "Decagon trains and serves its real-time voice AI models on Modal, runs production inference for its multi-model voice stack on Together AI's inference engine, and powers its support agents with OpenAI models.", "facts": [ { "tool": "Modal", "vendor": "Modal", "layer": "inference", "usage": "Trains and serves real-time voice AI models at scale.", "status": "current", "confidence": "high", "last_verified": "2026-07-27", "added": "2026-07-27", "evidence": [ { "url": "https://modal.com/blog/decagon-case-study", "type": "case-study", "date": "2025-11-13", "quote": "Modal’s infrastructure powered this progress, enabling Decagon to train and serve increasingly capable models" } ] }, { "tool": "Together AI", "vendor": "Together AI", "layer": "inference", "usage": "Runs low-latency production inference for the voice stack.", "status": "current", "confidence": "high", "last_verified": "2026-07-27", "added": "2026-07-27", "evidence": [ { "url": "https://www.together.ai/customers/decagon", "type": "case-study", "date": "2026-07", "quote": "Decagon uses Together’s inference engine as the execution layer" } ] }, { "tool": "OpenAI models", "vendor": "OpenAI", "layer": "models", "usage": "Powers AI agents for customer support and operations.", "status": "current", "confidence": "high", "last_verified": "2026-07-27", "added": "2026-07-27", "evidence": [ { "url": "https://openai.com/index/frontier-builders/", "type": "case-study", "date": "2026-07", "quote": "Decagon uses OpenAI models to power AI agents for customer support" } ] } ] }, { "slug": "deep-cogito", "name": "Deep Cogito", "website": "https://www.deepcogito.com", "what": "AI lab training open-weight hybrid reasoning models.", "profile": "Deep Cogito trains its large open-weight models on Together's customizable multi-node GPU clusters and serves the full Cogito model lineup on Together's dedicated inference infrastructure.", "facts": [ { "tool": "Together GPU Clusters", "vendor": "Together AI", "layer": "training", "usage": "Trains large open-weight models on multi-node GPU clusters.", "status": "current", "confidence": "high", "last_verified": "2026-07-27", "added": "2026-07-27", "evidence": [ { "url": "https://www.together.ai/customers/deep-cogito", "type": "case-study", "date": "2026-07", "quote": "Deep Cogito uses Together for customizable H100/H200 GPU clusters, reliable long-run training infrastructure" } ] }, { "tool": "Together AI", "vendor": "Together AI", "layer": "inference", "usage": "Hosts the full Cogito model lineup on dedicated infrastructure.", "status": "current", "confidence": "high", "last_verified": "2026-07-27", "added": "2026-07-27", "evidence": [ { "url": "https://www.together.ai/customers/deep-cogito", "type": "case-study", "date": "2026-07", "quote": "Together hosts the full Cogito model lineup, from 3B to 671B parameters, on dedicated inference infrastructure." } ] } ] }, { "slug": "deepgram", "name": "Deepgram", "website": "https://deepgram.com", "what": "Speech AI platform for real-time speech-to-text, text-to-speech, and voice agents.", "profile": "Deepgram rebuilt its own engineering around Claude — Claude Code is the default loop, reaching Slack, Asana, GitHub, and Grafana over MCP, and reusable solutions ship as shared skills in a company-wide agents repository. It cut over to Claude Enterprise in mid-2026 so retention, tool permissions, and MCP allowlists are centrally managed under its HIPAA and SOC 2 obligations.", "facts": [ { "tool": "Claude Code", "vendor": "Anthropic", "layer": "ai-coding", "usage": "Default engineering loop, connected to internal systems over MCP.", "status": "current", "confidence": "high", "last_verified": "2026-08-22", "added": "2026-08-22", "evidence": [ { "url": "https://claude.com/customers/deepgram", "type": "case-study", "date": "2026-08", "quote": "The default engineering loop runs through Claude Code" } ] }, { "tool": "Claude Enterprise", "vendor": "Anthropic", "layer": "internal-ai", "usage": "Company-wide deployment with SSO and centrally managed MCP allowlists.", "status": "current", "confidence": "high", "last_verified": "2026-08-22", "added": "2026-08-22", "evidence": [ { "url": "https://claude.com/customers/deepgram", "type": "case-study", "date": "2026-08", "quote": "Migrated the entire company to Claude Enterprise with SSO" } ] } ] }, { "slug": "delphi", "name": "Delphi", "website": "https://www.delphi.ai", "what": "Platform for creating interactive AI Digital Minds of real people.", "profile": "Delphi selected Pinecone to power agentic retrieval and vector search for every Digital Mind on its platform.", "facts": [ { "tool": "Pinecone", "vendor": "Pinecone", "layer": "rag-vector", "usage": "Powers agentic retrieval for conversational Digital Minds.", "status": "current", "confidence": "high", "last_verified": "2026-07-27", "added": "2026-07-27", "evidence": [ { "url": "https://www.pinecone.io/customers/delphi/", "type": "case-study", "date": "2026-07", "quote": "Delphi selected Pinecone to power agentic retrieval for every Digital Mind on their platform." } ] } ] }, { "slug": "docsbot", "name": "DocsBot", "website": "https://docsbot.ai", "what": "Platform for custom AI chatbots trained on documentation and content.", "profile": "DocsBot answers customer queries with Weaviate powering semantic and hybrid search at query time.", "facts": [ { "tool": "Weaviate", "vendor": "Weaviate", "layer": "rag-vector", "usage": "Powers semantic and hybrid search for customer queries.", "status": "current", "confidence": "high", "last_verified": "2026-07-27", "added": "2026-07-27", "evidence": [ { "url": "https://weaviate.io/case-studies/docsbot", "type": "case-study", "date": "2026-07", "quote": "At query time, Weaviate powers semantic and hybrid search, returning the most relevant context" } ] } ] }, { "slug": "dropbox", "name": "Dropbox", "website": "https://www.dropbox.com", "what": "File storage and collaboration platform, maker of Dropbox Dash.", "profile": "Dropbox builds multi-tier evaluation pipelines and catches production regressions with Braintrust as its evaluation flywheel.", "facts": [ { "tool": "Braintrust", "vendor": "Braintrust", "layer": "evals-observability", "usage": "Builds multi-tier evaluation pipelines and detects production regressions.", "status": "current", "confidence": "high", "last_verified": "2026-07-27", "added": "2026-07-27", "evidence": [ { "url": "https://www.braintrust.dev/customers/dropbox", "type": "case-study", "date": "2026-07", "quote": "Braintrust allows us to set up that flywheel." } ] } ] }, { "slug": "epilepsy-foundation", "name": "Epilepsy Foundation", "website": "https://www.epilepsy.com", "what": "US nonprofit supporting people with epilepsy and seizure disorders.", "profile": "The Epilepsy Foundation built Sage, a conversational AI epilepsy assistant grounded in vetted medical content and powered by Claude.", "facts": [ { "tool": "Claude", "vendor": "Anthropic", "layer": "models", "usage": "Powers Sage, a conversational epilepsy assistant grounded in vetted content.", "status": "current", "confidence": "high", "last_verified": "2026-07-27", "added": "2026-07-27", "evidence": [ { "url": "https://claude.com/customers/epilepsy-foundation", "type": "case-study", "date": "2026-07", "quote": "Sage itself is an AI epilepsy assistant: a conversational agent, built on Claude." }, { "url": "https://claude.com/customers/epilepsy-foundation-qa", "type": "case-study", "date": "2026-07", "quote": "Sage, is a Claude-powered AI epilepsy assistant" } ] } ] }, { "slug": "eve-legal", "name": "Eve", "website": "https://www.eve.legal", "what": "Legal AI platform for plaintiff law firms.", "profile": "Eve runs its heaviest workflows — analyzing firm operations, dissecting case files, drafting demand letters — on Claude, reached through the Claude Platform, Amazon Bedrock, and Google Cloud. It has also built Claude Code into every engineering workflow, including its internal developer tool Wall-E. Quality is measured in Braintrust against Plaintiff Bench, an expert-graded benchmark of plaintiff law tasks, with production failures reproduced as test cases and online scorers running against live traffic.", "facts": [ { "tool": "Claude", "vendor": "Anthropic", "layer": "models", "usage": "Runs case analysis and demand-letter drafting workflows.", "status": "current", "confidence": "high", "last_verified": "2026-07-27", "added": "2026-07-27", "evidence": [ { "url": "https://claude.com/customers/eve-legal", "type": "case-study", "date": "2026-07", "quote": "The harness reaches Claude through three routes, the native Claude Platform, Amazon Bedrock, and Google Cloud" } ] }, { "tool": "Claude Code", "vendor": "Anthropic", "layer": "ai-coding", "usage": "Built into every engineering workflow, including internal tool Wall-E.", "status": "current", "confidence": "high", "last_verified": "2026-07-27", "added": "2026-07-27", "evidence": [ { "url": "https://claude.com/customers/eve-legal", "type": "case-study", "date": "2026-07", "quote": "building Claude Code into every engineering workflow" } ] }, { "tool": "Braintrust", "vendor": "Braintrust", "layer": "evals-observability", "usage": "Measures agent quality against the Plaintiff Bench benchmark and production traffic.", "status": "current", "confidence": "high", "last_verified": "2026-08-22", "added": "2026-08-22", "evidence": [ { "url": "https://www.braintrust.dev/customers/eve", "type": "case-study", "date": "2026-08", "quote": "His job is making sure that work is good, and Braintrust is how he measures it." } ] } ] }, { "slug": "factory", "name": "Factory", "website": "https://factory.ai", "what": "AI software engineering agents (Droids) for enterprise development.", "profile": "Factory relies on Fireworks AI for reliable inference and day-zero access to newly released open models for its coding agents.", "facts": [ { "tool": "Fireworks AI", "vendor": "Fireworks AI", "layer": "inference", "usage": "Delivers reliable inference and rapid model access for agents.", "status": "current", "confidence": "high", "last_verified": "2026-07-27", "added": "2026-07-27", "evidence": [ { "url": "https://fireworks.ai/blog/Factory", "type": "case-study", "date": "2026-06-26", "quote": "Fireworks supports us by having these models available on basically day zero, typically well ahead of most other inference providers." } ] } ] }, { "slug": "figma", "name": "Figma", "website": "https://www.figma.com", "what": "Collaborative interface design platform behind Figma Design and Figma Make.", "profile": "Figma uses Claude to power Figma Make, which turns design mockups and prompts into interactive prototypes and working apps.", "facts": [ { "tool": "Claude", "vendor": "Anthropic", "layer": "models", "usage": "Powers Figma Make prototype and app generation from designs.", "status": "current", "confidence": "high", "last_verified": "2026-07-27", "added": "2026-07-27", "evidence": [ { "url": "https://www.claude.com/customers/figma", "type": "case-study", "date": "2026-07", "quote": "With Claude powering Figma Make, Figma enables teams to quickly transform ideas" } ] } ] }, { "slug": "fintool", "name": "Fintool", "website": "https://fintool.com", "what": "AI equity research copilot for institutional investors.", "profile": "Fintool benchmarks LLM output quality in real time with Braintrust, building repeatable evaluation workflows and automated quality checks for financial insights.", "facts": [ { "tool": "Braintrust", "vendor": "Braintrust", "layer": "evals-observability", "usage": "Benchmarks LLM output quality with repeatable evaluation workflows.", "status": "current", "confidence": "high", "last_verified": "2026-07-27", "added": "2026-07-27", "evidence": [ { "url": "https://www.braintrust.dev/customers/fintool", "type": "case-study", "date": "2026-07", "quote": "Fintool leverages Braintrust’s tools to benchmark the quality of LLM outputs in real time." } ] } ] }, { "slug": "genspark", "name": "Genspark", "website": "https://www.genspark.ai", "what": "AI agent company building autonomous research and productivity agents.", "profile": "Genspark trains large state-of-the-art open models with Fireworks' Reinforcement Fine-Tuning to power its Deep Research agent.", "facts": [ { "tool": "Fireworks AI", "vendor": "Fireworks AI", "layer": "training", "usage": "Trains large open models via reinforcement fine-tuning for Deep Research.", "status": "current", "confidence": "high", "last_verified": "2026-07-27", "added": "2026-07-27", "evidence": [ { "url": "https://fireworks.ai/blog/genspark", "type": "case-study", "date": "2025-10-31", "quote": "By leveraging Fireworks’ Reinforcement Fine Tuning to train large state-of-the-art open models" } ] } ] }, { "slug": "gong", "name": "Gong", "website": "https://www.gong.io", "what": "Revenue intelligence platform analyzing customer interactions.", "profile": "Gong selected Pinecone as the core database infrastructure behind Smart Trackers, storing and processing vector embeddings for concept tracking.", "facts": [ { "tool": "Pinecone", "vendor": "Pinecone", "layer": "rag-vector", "usage": "Stores vector embeddings for Smart Trackers concept tracking.", "status": "current", "confidence": "high", "last_verified": "2026-07-27", "added": "2026-07-27", "evidence": [ { "url": "https://www.pinecone.io/customers/gong/", "type": "case-study", "date": "2026-07", "quote": "Gong selected Pinecone as the core database infrastructure" } ] } ] }, { "slug": "graphite", "name": "Graphite", "website": "https://graphite.dev", "what": "AI-powered code review and stacked pull request platform.", "profile": "Graphite evaluates its AI code review features by running evaluations on annotated datasets in Braintrust and comparing model variants before deployment.", "facts": [ { "tool": "Braintrust", "vendor": "Braintrust", "layer": "evals-observability", "usage": "Evaluates code-review features and compares model variants.", "status": "current", "confidence": "high", "last_verified": "2026-07-27", "added": "2026-07-27", "evidence": [ { "url": "https://www.braintrust.dev/customers/graphite", "type": "case-study", "date": "2026-07", "quote": "run evaluations on both options using their annotated datasets in Braintrust" } ] } ] }, { "slug": "hubspot", "name": "HubSpot", "website": "https://www.hubspot.com", "what": "Customer platform for marketing, sales, and service teams.", "profile": "HubSpot identified Claude as a preferred solution in its AI tech stack and rolled it out across engineering, marketing, and customer success. Its engineers use Claude Code for development work, connected to internal infrastructure through MCP.", "facts": [ { "tool": "Claude", "vendor": "Anthropic", "layer": "internal-ai", "usage": "Supports engineering, marketing, and customer success workflows company-wide.", "status": "current", "confidence": "high", "last_verified": "2026-07-27", "added": "2026-07-27", "evidence": [ { "url": "https://claude.com/customers/hubspot", "type": "case-study", "date": "2026-07", "quote": "After evaluating multiple AI tools, HubSpot identified Claude as one of the preferred solutions in their AI tech stack." } ] }, { "tool": "Claude Code", "vendor": "Anthropic", "layer": "ai-coding", "usage": "Accelerates development, connected to internal infrastructure through MCP.", "status": "current", "confidence": "high", "last_verified": "2026-07-27", "added": "2026-07-27", "evidence": [ { "url": "https://claude.com/customers/hubspot", "type": "case-study", "date": "2026-07", "quote": "Engineering uses Claude Code for development work" } ] } ] }, { "slug": "innovative-solutions", "name": "Innovative Solutions", "website": "https://innovativesol.com", "what": "AWS premier partner delivering cloud and AI services.", "profile": "Innovative Solutions moved its DarcyIQ platform to Fireworks AI as its primary inference layer, powering multi-agent execution and model deployment.", "facts": [ { "tool": "Fireworks AI", "vendor": "Fireworks AI", "layer": "inference", "usage": "Primary inference layer for the DarcyIQ multi-agent platform.", "status": "current", "confidence": "high", "last_verified": "2026-07-27", "added": "2026-07-27", "evidence": [ { "url": "https://fireworks.ai/blog/innovative-solutions", "type": "case-study", "date": "2026-05-05", "quote": "the company moved its DarcyIQ platform to Fireworks AI as its primary inference layer." } ] } ] }, { "slug": "instabase", "name": "Instabase", "website": "https://instabase.com", "what": "AI platform for unstructured document data and automation.", "profile": "Instabase brought Weaviate into production across teams to index and retrieve unstructured data for enterprise AI applications.", "facts": [ { "tool": "Weaviate", "vendor": "Weaviate", "layer": "rag-vector", "usage": "Indexes and retrieves unstructured data for enterprise AI applications.", "status": "current", "confidence": "high", "last_verified": "2026-07-27", "added": "2026-07-27", "evidence": [ { "url": "https://weaviate.io/case-studies/instabase", "type": "case-study", "date": "2026-07", "quote": "teams across the organization worked together to bring Weaviate into production." } ] } ] }, { "slug": "jenova", "name": "Jenova", "website": "https://www.jenova.ai", "what": "Consumer AI agent platform routing across models and tools.", "profile": "Jenova runs agent memory and retrieval-augmented generation on Pinecone, which handles vector indexing and retrieval for its agent platform.", "facts": [ { "tool": "Pinecone", "vendor": "Pinecone", "layer": "rag-vector", "usage": "Provides vector indexing and retrieval for agent memory and RAG.", "status": "current", "confidence": "high", "last_verified": "2026-07-27", "added": "2026-07-27", "evidence": [ { "url": "https://www.pinecone.io/customers/jenova/", "type": "case-study", "date": "2026-07", "quote": "Pinecone handles vector indexing and retrieval." } ] } ] }, { "slug": "kapa", "name": "kapa.ai", "website": "https://www.kapa.ai", "what": "AI answer engine that turns product docs into technical support chatbots.", "profile": "kapa.ai maintains Weaviate as the core vector database and embeddings layer behind AI chatbots that answer technical product questions.", "facts": [ { "tool": "Weaviate", "vendor": "Weaviate", "layer": "rag-vector", "usage": "Core vector database and embeddings layer for technical chatbots.", "status": "current", "confidence": "high", "last_verified": "2026-07-27", "added": "2026-07-27", "evidence": [ { "url": "https://weaviate.io/case-studies/kapa", "type": "case-study", "date": "2026-07", "quote": "maintaining Weaviate as their core vector database and embeddings layer solution." } ] } ] }, { "slug": "klarna", "name": "Klarna", "website": "https://www.klarna.com", "what": "Swedish buy-now-pay-later payments and consumer banking company (NYSE: KLAR).", "profile": "Klarna is the poster child of enterprise AI adoption — and of its course corrections. It launched an OpenAI-powered customer-service assistant in February 2024 claiming the work of 700 full-time agents, consolidated 1,200+ SaaS tools (including Salesforce CRM) onto an internal Neo4j-based knowledge stack, then publicly walked back the AI-only support model in May 2025 and began rehiring humans for complex cases. As of late 2025 it still credits the assistant with 853 FTE-equivalents of work and $60M in savings, and in 2026 runs GenAI through a centralized in-house AI Gateway with LangGraph orchestration and LangSmith evals.", "facts": [ { "tool": "OpenAI models", "vendor": "OpenAI", "layer": "models", "usage": "Powers the customer-facing Klarna AI Assistant.", "status": "current", "confidence": "high", "last_verified": "2026-07-26", "added": "2026-07-26", "evidence": [ { "url": "https://www.klarna.com/international/press/klarna-ai-assistant-handles-two-thirds-of-customer-service-chats-in-its-first-month/", "type": "press", "date": "2024-02-27", "quote": "Klarna today announced its AI assistant powered by OpenAI." }, { "url": "https://www.customerexperiencedive.com/news/klarna-says-ai-agent-work-853-employees/805987/", "type": "press", "date": "2025-11-20", "quote": "We continue to see very demonstrable value from Klarna's AI assistant." } ] }, { "tool": "LangGraph", "vendor": "LangChain", "layer": "orchestration", "usage": "Agent framework the customer-service AI Assistant is built on.", "status": "current", "confidence": "high", "last_verified": "2026-07-26", "added": "2026-07-26", "evidence": [ { "url": "https://www.langchain.com/blog/customers-klarna", "type": "case-study", "date": "2025-02-12", "quote": "Built on LangGraph and powered by LangSmith, the AI Assistant handles tasks ranging from customer payments, to refunds" } ] }, { "tool": "LangSmith", "vendor": "LangChain", "layer": "evals-observability", "usage": "Provides tracing, evaluations, and prompt iteration for the AI Assistant.", "status": "current", "confidence": "high", "last_verified": "2026-07-26", "added": "2026-07-26", "evidence": [ { "url": "https://www.langchain.com/blog/customers-klarna", "type": "case-study", "date": "2025-02-12", "quote": "With LangSmith, Klarna could pinpoint what issues arose by seeing step-by-step how their AI assistant behaved." }, { "url": "https://www.jobsinforex.com/company/10686/klarna/", "type": "job-post", "date": "2026-07-21", "quote": "adapting and improving evaluation tooling for LLM applications using LangSmith" } ] }, { "tool": "Klarna AI Gateway", "vendor": "Klarna (in-house)", "layer": "inference", "usage": "Provides centralized low-latency model access for GenAI use cases company-wide.", "status": "current", "confidence": "high", "last_verified": "2026-07-26", "added": "2026-07-26", "evidence": [ { "url": "https://www.jobsinforex.com/company/10686/klarna/", "type": "job-post", "date": "2026-07-21", "quote": "manage and operate Klarna's AI Gateway, ensuring reliable, low-latency access to AI models at scale" } ] }, { "tool": "Neo4j", "vendor": "Neo4j", "layer": "rag-vector", "usage": "Graph database powering the internal knowledge stack behind Kiki.", "status": "current", "confidence": "medium", "last_verified": "2026-07-26", "added": "2026-07-26", "evidence": [ { "url": "https://www.cxtoday.com/crm/klarna-ceo-tremendously-embarrassed-by-salesforce-fallout-explains-what-really-happened/", "type": "press", "date": "2025-03-04", "quote": "we developed an internal tech stack, using Neo4j and other things, to start bringing data=knowledge together" }, { "url": "https://diginomica.com/those-shutting-down-salesforce-and-workday-rumors-klarna-no-we-didnt-replace-saas-llm-admits-ceo", "type": "press", "date": "2025-03-07", "quote": "Storing CRM data in an LLM would have its limitations." } ] }, { "tool": "GenAI governance tooling", "vendor": "Klarna (in-house)", "layer": "guardrails", "usage": "Gates GenAI use for safety and fintech compliance.", "status": "current", "practice": true, "confidence": "high", "last_verified": "2026-07-26", "added": "2026-07-26", "evidence": [ { "url": "https://www.jobsinforex.com/company/10686/klarna/", "type": "job-post", "date": "2026-07-21", "quote": "own tooling and processes that enable safe, compliant, and efficient GenAI use" } ] }, { "tool": "AI Assistant topic guardrails", "vendor": "Klarna (in-house)", "layer": "guardrails", "usage": "Restricts the assistant to supported topics with human handoff on sensitive queries.", "status": "current", "practice": true, "confidence": "medium", "last_verified": "2026-07-26", "added": "2026-07-26", "evidence": [ { "url": "https://blog.pragmaticengineer.com/klarnas-ai-chatbot/", "type": "press", "date": "2024-02-29", "quote": "When an \"out of bounds\" topic comes up, the chatbot hands over to a human agent" } ] }, { "tool": "ChatGPT Enterprise", "vendor": "OpenAI", "layer": "internal-ai", "usage": "Serves as the company-wide employee AI assistant.", "status": "current", "confidence": "high", "last_verified": "2026-07-26", "added": "2026-07-26", "evidence": [ { "url": "https://www.klarna.com/international/press/klarna-is-the-only-bank-among-chatgpt-enterprise-launch-customers/", "type": "press", "date": "2023-08-30", "quote": "ChatGPT Enterprise will give everyone at Klarna access to the power of OpenAI" }, { "url": "https://www.klarna.com/international/press/90-of-klarna-staff-are-using-ai-daily-game-changer-for-productivity/", "type": "press", "date": "2024-05-14", "quote": "Klarna's lawyers are using ChatGPT Enterprise" } ] }, { "tool": "Kiki", "vendor": "Klarna (in-house)", "layer": "internal-ai", "usage": "Internal knowledge assistant answering employee questions daily.", "status": "current", "confidence": "high", "last_verified": "2026-07-26", "added": "2026-07-26", "evidence": [ { "url": "https://www.klarna.com/international/press/90-of-klarna-staff-are-using-ai-daily-game-changer-for-productivity/", "type": "press", "date": "2024-05-14", "quote": "Kiki, Klarna's bespoke internal AI assistant, which has adeptly responded to over 250,000 inquiries" } ] }, { "tool": "Cursor", "vendor": "Anysphere", "layer": "ai-coding", "usage": "The CEO vibe-codes feature prototypes before handing them to engineering.", "status": "current", "confidence": "medium", "last_verified": "2026-07-26", "added": "2026-07-26", "evidence": [ { "url": "https://finance.yahoo.com/news/klarnas-ceo-says-uses-vibe-052314552.html", "type": "press", "date": "2025-09-15", "quote": "I was a business person, and then I just started exploring Cursor" }, { "url": "https://www.sourcery.vc/p/klarna-ceo-sebastian-siemiatkowski", "type": "podcast", "date": "2025-09-12", "quote": "I've been vibe coding my whole life." } ] }, { "tool": "Salesforce", "vendor": "Salesforce", "layer": "internal-ai", "usage": "CRM retired during Klarna's AI-driven SaaS consolidation.", "status": "historical", "confidence": "medium", "last_verified": "2026-07-26", "added": "2026-07-26", "until": "2025-03", "evidence": [ { "url": "https://www.cxtoday.com/crm/klarna-ceo-tremendously-embarrassed-by-salesforce-fallout-explains-what-really-happened/", "type": "press", "date": "2025-03-04", "quote": "The side consequence of this was the liquidation of SaaS – not all of them, but a lot of them." }, { "url": "https://diginomica.com/those-shutting-down-salesforce-and-workday-rumors-klarna-no-we-didnt-replace-saas-llm-admits-ceo", "type": "press", "date": "2025-03-07", "quote": "So no, we did not replace SaaS with an LLM" } ] }, { "tool": "AI-only customer support (Klarna AI Assistant)", "vendor": "Klarna (in-house)", "layer": "internal-ai", "usage": "Fully AI-run support strategy later reversed in favor of a hybrid model.", "status": "historical", "practice": true, "confidence": "medium", "last_verified": "2026-07-26", "added": "2026-07-26", "until": "2025-05", "evidence": [ { "url": "https://www.customerexperiencedive.com/news/klarna-reinvests-human-talent-customer-service-AI-chatbot/747586/", "type": "press", "date": "2025-05-09", "quote": "Really investing in the quality of the human support is the way of the future for us." } ] } ] }, { "slug": "latent-health", "name": "Latent Health", "website": "https://latenthealth.com", "what": "Clinical AI company for pharmacy intelligence.", "profile": "Latent Health trains its clinical AI models — using multi-node reinforcement learning and long-context training — on Together Instant Clusters.", "facts": [ { "tool": "Together GPU Clusters", "vendor": "Together AI", "layer": "training", "usage": "Trains clinical models with multi-node RL and long-context runs.", "status": "current", "confidence": "high", "last_verified": "2026-07-27", "added": "2026-07-27", "evidence": [ { "url": "https://www.together.ai/customers/latent-health", "type": "case-study", "date": "2026-07", "quote": "Latent Health chose Together Instant Clusters for clinical AI training" } ] } ] }, { "slug": "league", "name": "League", "website": "https://league.com", "what": "Digital health platform powering consumer health experiences for health plans and providers.", "profile": "League signed its Claude Enterprise agreement and had the whole company live within days, then spent about a quarter building the controls its regulated environment needs — patient data stays in an environment Claude cannot reach. Engineering runs larger builds through Swarm, an internal orchestration tool that executes overnight, while finance, security, and design teams automate their own processes.", "facts": [ { "tool": "Claude Enterprise", "vendor": "Anthropic", "layer": "internal-ai", "usage": "Company-wide deployment; finance, security, and design teams automate their own work.", "status": "current", "confidence": "high", "last_verified": "2026-08-22", "added": "2026-08-22", "evidence": [ { "url": "https://claude.com/customers/league", "type": "case-study", "date": "2026-08", "quote": "League signed its Claude Enterprise agreement on a Friday and had the whole company live by Monday" }, { "url": "https://claude.com/customers/league", "type": "case-study", "date": "2026-08", "quote": "League cuts product development cycle times in half with Claude" } ] }, { "tool": "Claude Code", "vendor": "Anthropic", "layer": "ai-coding", "usage": "Underpins Swarm, the internal orchestrator running overnight autonomous coding sessions.", "status": "current", "confidence": "high", "last_verified": "2026-08-22", "added": "2026-08-22", "evidence": [ { "url": "https://claude.com/customers/league", "type": "case-study", "date": "2026-08", "quote": "Runs overnight autonomous coding sessions through Swarm, League's internal orchestration tool built on Claude Code" } ] } ] }, { "slug": "lg-cns", "name": "LG CNS", "website": "https://www.lgcns.com", "what": "Korean IT services company of LG Group delivering enterprise system integration.", "profile": "LG CNS adopted Claude Code as the standard development approach across its 200-engineer Build Center and relies on Claude Opus for architecture design, legacy analysis, and complex code generation in enterprise modernization work.", "facts": [ { "tool": "Claude Code", "vendor": "Anthropic", "layer": "ai-coding", "usage": "Standard development approach across the enterprise modernization Build Center.", "status": "current", "confidence": "high", "last_verified": "2026-07-27", "added": "2026-07-27", "evidence": [ { "url": "https://claude.com/customers/lg-cns", "type": "case-study", "date": "2026-07", "quote": "Adopted Claude Code as the standard development approach across the 200-engineer Build Center" } ] }, { "tool": "Claude", "vendor": "Anthropic", "layer": "models", "usage": "Drives architecture design, legacy analysis, and complex code generation.", "status": "current", "confidence": "high", "last_verified": "2026-07-27", "added": "2026-07-27", "evidence": [ { "url": "https://claude.com/customers/lg-cns", "type": "case-study", "date": "2026-07", "quote": "The team relies on Claude Opus 4.6 for core work" } ] } ] }, { "slug": "loom", "name": "Loom", "website": "https://www.loom.com", "what": "Async video messaging platform, part of Atlassian.", "profile": "Loom measures AI output quality by running evals with custom scoring functions on Braintrust.", "facts": [ { "tool": "Braintrust", "vendor": "Braintrust", "layer": "evals-observability", "usage": "Runs evals with custom scoring functions on AI output quality.", "status": "current", "confidence": "high", "last_verified": "2026-07-27", "added": "2026-07-27", "evidence": [ { "url": "https://www.braintrust.dev/customers/loom", "type": "case-study", "date": "2026-07", "quote": "To answer that question, they started running evals on Braintrust with their own custom scoring functions." } ] } ] }, { "slug": "loti-ai", "name": "Loti AI", "website": "https://goloti.com", "what": "Digital likeness protection platform for public figures.", "profile": "Loti AI partners with Weaviate to power its digital identity protection platform, scanning for and removing unauthorized content.", "facts": [ { "tool": "Weaviate", "vendor": "Weaviate", "layer": "rag-vector", "usage": "Powers likeness detection and unauthorized-content removal at scale.", "status": "current", "confidence": "high", "last_verified": "2026-07-27", "added": "2026-07-27", "evidence": [ { "url": "https://weaviate.io/case-studies/loti", "type": "case-study", "date": "2026-07", "quote": "Loti AI partnered with Weaviate to power their digital identity protection platform" } ] } ] }, { "slug": "lovable", "name": "Lovable", "website": "https://lovable.dev", "what": "AI app builder turning prompts into full applications.", "profile": "Every app generation session in Lovable is served by Modal Sandboxes, which run LLM-generated code in safe, isolated environments.", "facts": [ { "tool": "Modal Sandboxes", "vendor": "Modal", "layer": "guardrails", "usage": "Runs LLM-generated app code in safe, isolated sandboxes.", "status": "current", "confidence": "high", "last_verified": "2026-07-27", "added": "2026-07-27", "evidence": [ { "url": "https://modal.com/blog/lovable-case-study", "type": "case-study", "date": "2025-07-07", "quote": "Modal Sandboxes are now used to serve every app generation session in Lovable." } ] } ] }, { "slug": "marvelx", "name": "MarvelX", "website": "https://marvelx.ai", "what": "Agentic AI platform for insurance claims automation.", "profile": "MarvelX uses Weaviate as the foundation of its autonomous agent platform for AI-driven claims automation and multimodal search.", "facts": [ { "tool": "Weaviate", "vendor": "Weaviate", "layer": "rag-vector", "usage": "Foundation for claims-automation agents and multimodal search.", "status": "current", "confidence": "high", "last_verified": "2026-07-27", "added": "2026-07-27", "evidence": [ { "url": "https://weaviate.io/case-studies/marvelx", "type": "case-study", "date": "2026-07", "quote": "MarvelX AI uses Weaviate as a foundation for their autonomous agent platform" } ] } ] }, { "slug": "melange", "name": "Melange", "website": "https://melange.com", "what": "Innovation intelligence platform for patent and research search.", "profile": "Melange indexes massive patent and academic datasets for semantic search, with its entire search operation relying on Pinecone as the first retrieval step.", "facts": [ { "tool": "Pinecone", "vendor": "Pinecone", "layer": "rag-vector", "usage": "Indexes patent and academic datasets for semantic search.", "status": "current", "confidence": "high", "last_verified": "2026-07-27", "added": "2026-07-27", "evidence": [ { "url": "https://www.pinecone.io/customers/melange/", "type": "case-study", "date": "2026-07", "quote": "Our entire search operation relies on Pinecone as the first step in the process." } ] } ] }, { "slug": "metabuddy", "name": "MetaBuddy", "website": "https://metabuddy.app", "what": "AI wellness app unifying health and fitness data.", "profile": "MetaBuddy integrated a Weaviate vector database with Weaviate's Query Agent to deliver semantic search and personalized AI coaching over unified wellness data.", "facts": [ { "tool": "Weaviate", "vendor": "Weaviate", "layer": "rag-vector", "usage": "Provides semantic search over unified wellness data.", "status": "current", "confidence": "high", "last_verified": "2026-07-27", "added": "2026-07-27", "evidence": [ { "url": "https://weaviate.io/case-studies/metabuddy", "type": "case-study", "date": "2026-07", "quote": "MetaBuddy integrated a Weaviate vector database with Query Agent to deliver semantic search" } ] }, { "tool": "Weaviate Query Agent", "vendor": "Weaviate", "layer": "orchestration", "usage": "Orchestrates intelligent queries for personalized coaching and insights.", "status": "current", "confidence": "high", "last_verified": "2026-07-27", "added": "2026-07-27", "evidence": [ { "url": "https://weaviate.io/case-studies/metabuddy", "type": "case-study", "date": "2026-07", "quote": "Weaviate’s Query Agent allowed us to unify our users’ wellness data into a single intelligent interface." } ] } ] }, { "slug": "miro", "name": "Miro", "website": "https://miro.com", "what": "Collaborative visual workspace for distributed teams.", "profile": "Miro's spread of Claude Cowork came from non-engineering champions rather than a rollout plan — an influencer content reviewer that checks scripts against an event brief, a scheduled Friday task that pulls context from Slack, email, and calendar and writes updates back to a Miro board, and a thought partner for early-stage problems. Quality standards live in shared skills that travel with the work, and a human owns every external call.", "facts": [ { "tool": "Claude Cowork", "vendor": "Anthropic", "layer": "internal-ai", "usage": "Automates influencer content review and weekly project tracking for marketing teams.", "status": "current", "confidence": "high", "last_verified": "2026-08-22", "added": "2026-08-22", "evidence": [ { "url": "https://claude.com/customers/miro-qa", "type": "case-study", "date": "2026-08", "quote": "Cowork took over the busywork of documenting the past so I can focus entirely on strategizing for the future." } ] } ] }, { "slug": "monday-com", "name": "monday.com", "website": "https://monday.com", "what": "Work OS platform for project and service management.", "profile": "monday.com builds its AI service workforce as customizable LangGraph-based ReAct agents and runs a code-first evaluation strategy on LangSmith using the Vitest integration.", "facts": [ { "tool": "LangSmith", "vendor": "LangChain", "layer": "evals-observability", "usage": "Drives a code-first evaluation strategy for AI agents.", "status": "current", "confidence": "high", "last_verified": "2026-07-27", "added": "2026-07-27", "evidence": [ { "url": "https://blog.langchain.com/customers-monday/", "type": "case-study", "date": "2026-02-18", "quote": "we utilized the LangSmith Vitest integration ." } ] }, { "tool": "LangGraph", "vendor": "LangChain", "layer": "orchestration", "usage": "Underpins customizable ReAct agents for enterprise service management.", "status": "current", "confidence": "high", "last_verified": "2026-07-27", "added": "2026-07-27", "evidence": [ { "url": "https://blog.langchain.com/customers-monday/", "type": "case-study", "date": "2026-02-18", "quote": "AI service workforce is a customizable, LangGraph-based, ReAct agent" } ] } ] }, { "slug": "morningstar", "name": "Morningstar", "website": "https://www.morningstar.com", "what": "Investment research and financial data company.", "profile": "Morningstar built its Intelligence Engine Platform on Weaviate's vector database, powering a trustworthy AI-driven financial research assistant and low-latency search through its Corpus API.", "facts": [ { "tool": "Weaviate", "vendor": "Weaviate", "layer": "rag-vector", "usage": "Powers the Intelligence Engine behind AI financial research.", "status": "current", "confidence": "high", "last_verified": "2026-07-27", "added": "2026-07-27", "evidence": [ { "url": "https://weaviate.io/case-studies/morningstar", "type": "case-study", "date": "2026-07", "quote": "Through our Corpus API connected to Weaviate, users can build very powerful, low latency search engines in minutes" } ] } ] }, { "slug": "national-domestic-workers-alliance", "name": "National Domestic Workers Alliance", "website": "https://www.domesticworkers.org", "what": "US nonprofit organizing and advocating for domestic workers.", "profile": "NDWA built Ask Aya, an AI support tool on Claude that helps domestic workers self-advocate and is governed by the workers it serves.", "facts": [ { "tool": "Claude", "vendor": "Anthropic", "layer": "models", "usage": "Powers Ask Aya, a worker-governed AI support tool.", "status": "current", "confidence": "high", "last_verified": "2026-07-27", "added": "2026-07-27", "evidence": [ { "url": "https://claude.com/customers/national-domestic-workers-alliance-qa", "type": "case-study", "date": "2026-07", "quote": "NDWA built Ask Aya, an AI support tool built on Claude and governed by the workers it serves." } ] } ] }, { "slug": "navan", "name": "Navan", "website": "https://navan.com", "what": "Corporate travel and expense management platform.", "profile": "Navan partnered with Braintrust to build the evaluation loop for its AI voice agent, logging reasoning steps to monitor quality.", "facts": [ { "tool": "Braintrust", "vendor": "Braintrust", "layer": "evals-observability", "usage": "Runs the evaluation loop for AI voice agent quality.", "status": "current", "confidence": "high", "last_verified": "2026-07-27", "added": "2026-07-27", "evidence": [ { "url": "https://www.braintrust.dev/customers/navan", "type": "case-study", "date": "2026-07", "quote": "This is where Navan partnered with Braintrust to build their evaluation loop." } ] } ] }, { "slug": "neople", "name": "Neople", "website": "https://neople.io", "what": "AI digital coworkers for customer support teams.", "profile": "Neople uses Weaviate as its vector database of choice for reliable, accurate, and fast querying of company-specific knowledge.", "facts": [ { "tool": "Weaviate", "vendor": "Weaviate", "layer": "rag-vector", "usage": "Serves company-specific knowledge with reliable, fast vector querying.", "status": "current", "confidence": "high", "last_verified": "2026-07-27", "added": "2026-07-27", "evidence": [ { "url": "https://weaviate.io/case-studies/neople", "type": "case-study", "date": "2026-07", "quote": "Weaviate as their vector database of choice for reliable, accurate and fast querying" } ] } ] }, { "slug": "notion", "name": "Notion", "website": "https://www.notion.com", "what": "Connected AI workspace for docs, wikis, projects, and enterprise search (Notion AI, Agents, Q&A).", "profile": "Notion ships one of the most-watched applied-AI stacks in SaaS: a deliberately model-agnostic layer over OpenAI, Anthropic, Google, and xAI frontier models (users pick the model per agent), on top of a heavily custom in-house agent harness rebuilt four-plus times. They buy best-of-breed infrastructure — turbopuffer for vectors, Braintrust for evals, Fireworks for fine-tuned model serving, Anyscale-managed Ray for embeddings — while keeping orchestration, tools, and retrieval logic in-house. Their subprocessor list is unusually candid, also naming Cerebras, Baseten, Exa, Parallel Web Systems, and MCP security vendor Runlayer.", "facts": [ { "tool": "Claude", "vendor": "Anthropic", "layer": "models", "usage": "Core LLM provider for Notion AI and long-running agent workflows.", "status": "current", "confidence": "high", "last_verified": "2026-07-26", "added": "2026-07-26", "evidence": [ { "url": "https://www.notion.com/releases/2025-09-18", "type": "docs", "date": "2025-09-18", "quote": "The latest—like Claude Sonnet 4 and GPT-5—are already built in" }, { "url": "https://claude.com/customers/notion", "type": "case-study", "date": "2026-07", "quote": "Opus 4.6 excels at interpreting what users actually want, producing shareable content on the first try" }, { "url": "https://www.notion.com/help/notion-ai-security-practices", "type": "docs", "date": "2026-07", "quote": "hosted by Notion as well as by organizations such as Anthropic and OpenAI" } ] }, { "tool": "GPT-5", "vendor": "OpenAI", "layer": "models", "usage": "Frontier OpenAI models offered in the agent model picker.", "status": "current", "confidence": "high", "last_verified": "2026-07-26", "added": "2026-07-26", "evidence": [ { "url": "https://www.notion.com/releases/2025-09-18", "type": "docs", "date": "2025-09-18", "quote": "The latest—like Claude Sonnet 4 and GPT-5—are already built in" }, { "url": "https://registora.com/providers/notion", "type": "subprocessor", "date": "2026-07-09", "quote": "Service provider for hosting large language models and embeddings and for abuse prevention" } ] }, { "tool": "Gemini", "vendor": "Google", "layer": "models", "usage": "Google models offered in the multi-model agent picker.", "status": "current", "confidence": "medium", "last_verified": "2026-07-26", "added": "2026-07-26", "evidence": [ { "url": "https://registora.com/providers/notion", "type": "subprocessor", "date": "2026-07-09", "quote": "Service provider for hosting large language models and embeddings" } ] }, { "tool": "Fireworks AI", "vendor": "Fireworks AI", "layer": "inference", "usage": "Hosts and serves fine-tuned and open-weight models for low-latency features.", "status": "current", "confidence": "high", "last_verified": "2026-07-26", "added": "2026-07-26", "evidence": [ { "url": "https://fireworks.ai/blog/Story-Notion", "type": "case-study", "date": "2025-07-25", "quote": "By fine-tuning models, we reduced latency from about 2 seconds to 350 milliseconds" }, { "url": "https://registora.com/providers/notion", "type": "subprocessor", "date": "2026-07-09", "quote": "Service provider for hosting large language models and embeddings" } ] }, { "tool": "Anyscale (Ray)", "vendor": "Anyscale", "layer": "inference", "usage": "Runs the near-real-time embeddings indexing pipeline on managed Ray.", "status": "current", "confidence": "high", "last_verified": "2026-07-26", "added": "2026-07-26", "evidence": [ { "url": "https://www.notion.com/blog/two-years-of-vector-search-at-notion", "type": "eng-blog", "date": "2026-02-19", "quote": "we set out to migrating our near real-time embeddings pipeline to Ray running on Anyscale" }, { "url": "https://www.notion.com/blog/two-years-of-vector-search-at-notion", "type": "eng-blog", "date": "2026-02-19", "quote": "Ray lets us run open-source embedding models directly, without being gated by external providers" }, { "url": "https://www.anyscale.com/resources/case-study/notion", "type": "case-study", "date": "2026-07", "quote": "pull a model from Hugging Face and run it ourselves on-prem" } ] }, { "tool": "Custom agent harness", "vendor": "Notion", "layer": "orchestration", "usage": "In-house agent framework behind Notion Agents with extensive internal tools.", "status": "current", "practice": true, "confidence": "high", "last_verified": "2026-07-26", "added": "2026-07-26", "evidence": [ { "url": "https://www.latent.space/p/notion", "type": "podcast", "date": "2026-04-15", "quote": "There's now like over a hundred tools. Just for all, all the crazy notion stuff." }, { "url": "https://www.latent.space/p/notion", "type": "podcast", "date": "2026-04-15", "quote": "We have rebuilt our harness three or four times." } ] }, { "tool": "Notion MCP server", "vendor": "Notion", "layer": "orchestration", "usage": "Official hosted MCP server letting external AI clients act on workspaces.", "status": "current", "confidence": "high", "last_verified": "2026-07-26", "added": "2026-07-26", "evidence": [ { "url": "https://www.notion.com/blog/notions-hosted-mcp-server-an-inside-look", "type": "eng-blog", "date": "2025-07-15", "quote": "manages sessions and securely stores the API token from the OAuth exchange" }, { "url": "https://github.com/makenotion/notion-mcp-server", "type": "github", "date": "2026-07", "quote": "Official Notion MCP Server" } ] }, { "tool": "Claude Managed Agents", "vendor": "Anthropic", "layer": "orchestration", "usage": "Embeds Anthropic-hosted coding agents inside Notion via the Managed Agents API.", "status": "current", "confidence": "high", "last_verified": "2026-07-27", "added": "2026-07-27", "evidence": [ { "url": "https://claude.com/customers/notion-qa", "type": "case-study", "date": "2026-07", "quote": "Managed Agents was great because we just pull in the API and it works within the product." } ] }, { "tool": "turbopuffer", "vendor": "turbopuffer", "layer": "rag-vector", "usage": "Primary vector database for Notion AI search and Q&A.", "status": "current", "confidence": "high", "last_verified": "2026-07-26", "added": "2026-07-26", "evidence": [ { "url": "https://www.notion.com/blog/two-years-of-vector-search-at-notion", "type": "eng-blog", "date": "2026-02-19", "quote": "we committed to migrating our entire multi billion object workload to turbopuffer in late 2024" }, { "url": "https://www.notion.com/help/notion-ai-security-practices", "type": "docs", "date": "2026-07", "quote": "stores it in a vector database (e.g., Turbopuffer)" }, { "url": "https://registora.com/providers/notion", "type": "subprocessor", "date": "2026-07-09", "quote": "Vector database for storing embeddings" } ] }, { "tool": "OpenAI embeddings API", "vendor": "OpenAI", "layer": "rag-vector", "usage": "Embeds workspace pages for Q&A retrieval.", "status": "current", "confidence": "high", "last_verified": "2026-07-26", "added": "2026-07-26", "evidence": [ { "url": "https://www.notion.com/help/notion-ai-security-practices", "type": "docs", "date": "2026-07", "quote": "we generate an embedding by using an OpenAI zero-retention embeddings API" } ] }, { "tool": "Apache Kafka", "vendor": "Apache", "layer": "rag-vector", "usage": "Streams page edits into the embedding-indexing pipeline.", "status": "current", "confidence": "high", "last_verified": "2026-07-26", "added": "2026-07-26", "evidence": [ { "url": "https://www.notion.com/blog/two-years-of-vector-search-at-notion", "type": "eng-blog", "date": "2026-02-19", "quote": "Real-time updates via Kafka consumers that process individual page edits as they happen" } ] }, { "tool": "Exa", "vendor": "Exa", "layer": "rag-vector", "usage": "Provides live web results for AI research and search features.", "status": "current", "confidence": "medium", "last_verified": "2026-07-26", "added": "2026-07-26", "evidence": [ { "url": "https://registora.com/providers/notion", "type": "subprocessor", "date": "2026-07-09", "quote": "Service provider for providing web search results" } ] }, { "tool": "Braintrust", "vendor": "Braintrust", "layer": "evals-observability", "usage": "Eval and LLM-observability platform for regression and frontier-model testing.", "status": "current", "confidence": "high", "last_verified": "2026-07-26", "added": "2026-07-26", "evidence": [ { "url": "https://www.braintrust.dev/customers/notion", "type": "case-study", "date": "2026-07", "quote": "I sat down in Braintrust and looked at some of the worst experiences our customers had" } ] }, { "tool": "Custom fine-tuned models", "vendor": "Notion", "layer": "training", "usage": "Fine-tunes small open-weight models for search, routing, and function calling.", "status": "current", "practice": true, "confidence": "high", "last_verified": "2026-07-26", "added": "2026-07-26", "evidence": [ { "url": "https://fireworks.ai/blog/Story-Notion", "type": "case-study", "date": "2025-07-25", "quote": "By fine-tuning models, we reduced latency from about 2 seconds to 350 milliseconds" }, { "url": "https://www.latent.space/p/notion", "type": "podcast", "date": "2026-04-15", "quote": "Our, um, fine tuned and open source models are served on GPUs, right?" } ] }, { "tool": "AI LEAP Program", "vendor": "Notion", "layer": "data-labeling", "usage": "Opt-in program feeding customer workspace data into model improvement.", "status": "current", "practice": true, "confidence": "high", "last_verified": "2026-07-26", "added": "2026-07-26", "evidence": [ { "url": "https://www.notion.com/help/ai-safety", "type": "docs", "date": "2026-07", "quote": "allows for sharing of workspace data to improve underlying models" } ] }, { "tool": "Zero-retention LLM contracts", "vendor": "Notion", "layer": "guardrails", "usage": "Enforces zero-retention and no-training contracts with AI subprocessors.", "status": "current", "practice": true, "confidence": "high", "last_verified": "2026-07-26", "added": "2026-07-26", "evidence": [ { "url": "https://www.notion.com/help/notion-ai-security-practices", "type": "docs", "date": "2026-07", "quote": "our LLM providers utilize zero data retention for Enterprise plan workspaces" }, { "url": "https://www.notion.com/help/notion-ai-security-practices", "type": "docs", "date": "2026-07", "quote": "prohibit the use of Customer Data to train their models" } ] }, { "tool": "Vercel Sandbox", "vendor": "Vercel", "layer": "guardrails", "usage": "Runs untrusted Notion Workers code in isolated sandboxes.", "status": "current", "confidence": "high", "last_verified": "2026-07-27", "added": "2026-07-27", "evidence": [ { "url": "https://vercel.com/customers/notion-workers-vercel-sandbox", "type": "case-study", "date": "2026-03-12", "quote": "Under the hood, every Worker runs on Vercel Sandbox ." } ] }, { "tool": "Runlayer", "vendor": "Anysource (Runlayer)", "layer": "guardrails", "usage": "Secures Notion's MCP connections and agent tool traffic.", "status": "current", "confidence": "medium", "last_verified": "2026-07-26", "added": "2026-07-26", "evidence": [ { "url": "https://registora.com/providers/notion", "type": "subprocessor", "date": "2026-07-09", "quote": "Security for MCP connections" } ] }, { "tool": "Claude Code", "vendor": "Anthropic", "layer": "ai-coding", "usage": "Fastest-growing agentic coding assistant among Notion's engineers.", "status": "current", "confidence": "medium", "last_verified": "2026-07-26", "added": "2026-07-26", "evidence": [ { "url": "https://www.theinformation.com/newsletters/ai-agenda/notion-switching-cursor-claude-code-codex", "type": "press", "date": "2026-03-24", "quote": "Claude Code and Codex Are Outpacing Cursor Among Notion's Engineers" } ] }, { "tool": "Cursor", "vendor": "Anysphere", "layer": "ai-coding", "usage": "AI IDE in active use by Notion engineers.", "status": "current", "confidence": "medium", "last_verified": "2026-07-26", "added": "2026-07-26", "evidence": [ { "url": "https://www.anyscale.com/resources/case-study/notion", "type": "case-study", "date": "2026-07", "quote": "Anyscale lets me use my preferred tools like VSCode or Cursor on my laptop" }, { "url": "https://www.theinformation.com/newsletters/ai-agenda/notion-switching-cursor-claude-code-codex", "type": "press", "date": "2026-03-24", "quote": "Claude Code and Codex Are Outpacing Cursor Among Notion's Engineers" } ] }, { "tool": "Notion AI Agents (dogfooding)", "vendor": "Notion", "layer": "internal-ai", "usage": "Employees run the company on its own AI agents before public release.", "status": "current", "practice": true, "confidence": "high", "last_verified": "2026-07-26", "added": "2026-07-26", "evidence": [ { "url": "https://www.latent.space/p/notion", "type": "podcast", "date": "2026-04-15", "quote": "No one uses Notion in their job as much as people that work at Notion." }, { "url": "https://www.latent.space/p/notion", "type": "podcast", "date": "2026-04-15", "quote": "Everyone is using the same instance of notion with like a lot of flags on for these prototypes people build" } ] } ] }, { "slug": "obviant", "name": "Obviant", "website": "https://www.obviant.com", "what": "Defense acquisition data and intelligence platform.", "profile": "Obviant powers hybrid search over defense acquisition data with Pinecone to surface recommendations quickly.", "facts": [ { "tool": "Pinecone", "vendor": "Pinecone", "layer": "rag-vector", "usage": "Powers hybrid search for defense acquisition recommendations.", "status": "current", "confidence": "high", "last_verified": "2026-07-27", "added": "2026-07-27", "evidence": [ { "url": "https://www.pinecone.io/customers/obviant/", "type": "case-study", "date": "2026-07", "quote": "Pinecone enables that by helping us cut through the noise, fast." } ] } ] }, { "slug": "perplexity", "name": "Perplexity", "website": "https://www.perplexity.ai", "what": "AI answer engine combining a proprietary web index with LLMs to deliver cited, conversational search.", "profile": "Perplexity is the highest-profile AI-native search company: it runs its own crawler and 200B+ URL index on Vespa, post-trains its own Sonar models on open-weights Llama bases, and serves them on a largely self-built GPU inference stack (Rust/CUDA engine, open-sourced RDMA kernels) spread across AWS and other clouds. At the same time it is pragmatically multi-vendor at the model layer, reselling GPT, Claude, Gemini and Grok in its Pro model picker, and it publishes an unusual amount about its infrastructure via its research blog, vendor case studies, and job postings.", "facts": [ { "tool": "Sonar", "vendor": "Perplexity", "layer": "models", "usage": "In-house search model family powering default search.", "status": "current", "confidence": "high", "last_verified": "2026-07-26", "added": "2026-07-26", "evidence": [ { "url": "https://www.cerebras.ai/press-release/cerebras-powers-perplexity-sonar-with-industrys-fastest-ai-inference", "type": "press", "date": "2025-02-11", "quote": "Built on the robust foundation of Llama 3.3 70B, Sonar represents a significant advancement in answer quality, factuality, and readability" } ] }, { "tool": "Llama", "vendor": "Meta", "layer": "models", "usage": "Open-weights base models served in production and underpinning Sonar.", "status": "current", "confidence": "high", "last_verified": "2026-07-26", "added": "2026-07-26", "evidence": [ { "url": "https://developer.nvidia.com/blog/spotlight-perplexity-ai-serves-400-million-search-queries-a-month-using-nvidia-inference-stack", "type": "case-study", "date": "2024-12-05", "quote": "different variations of the popular open source Llama 3.1 models like 8B, 70B, and 405B" }, { "url": "https://www.cerebras.ai/press-release/cerebras-powers-perplexity-sonar-with-industrys-fastest-ai-inference", "type": "press", "date": "2025-02-11", "quote": "Built on the robust foundation of Llama 3.3 70B" } ] }, { "tool": "Claude", "vendor": "Anthropic", "layer": "models", "usage": "Offered in free and paid tiers through Amazon Bedrock.", "status": "current", "confidence": "high", "last_verified": "2026-07-26", "added": "2026-07-26", "evidence": [ { "url": "https://claude.com/customers/perplexity", "type": "case-study", "date": "2024-07", "quote": "Perplexity uses Amazon Bedrock to serve Claude models on both the free and paid versions of its search engine" } ] }, { "tool": "GPT-5", "vendor": "OpenAI", "layer": "models", "usage": "Flagship OpenAI models offered in the paid model picker.", "status": "current", "confidence": "high", "last_verified": "2026-07-26", "added": "2026-07-26", "evidence": [ { "url": "https://x.com/perplexity_ai/status/1953537170964459632", "type": "social", "date": "2025-08-07", "quote": "GPT-5 is now available on Perplexity and Comet for Max and Pro subscribers." } ] }, { "tool": "R1-1776", "vendor": "Perplexity", "layer": "models", "usage": "Post-trained open-weights reasoning model with censorship removed.", "status": "current", "confidence": "medium", "last_verified": "2026-07-26", "added": "2026-07-26", "evidence": [ { "url": "https://www.eweek.com/news/perplexity-ai-deepseek-r1-post-training/", "type": "press", "date": "2025-02-20", "quote": "We are not able to make use of R1's powerful reasoning capabilities without first mitigating its bias and censorship" } ] }, { "tool": "In-house inference engine", "vendor": "Perplexity", "layer": "inference", "usage": "Custom Rust and CUDA engine serving every query across a multi-cloud GPU fleet.", "status": "current", "practice": true, "confidence": "high", "last_verified": "2026-07-26", "added": "2026-07-26", "evidence": [ { "url": "https://jobs.ashbyhq.com/perplexity/8a976851-9bef-4b07-8d36-567fa9540aef", "type": "job-post", "date": "2026-04-13", "quote": "We build and run the inference engine behind every Perplexity query and deploy dozens of model architectures at scale" }, { "url": "https://jobs.ashbyhq.com/perplexity/4b8347e9-615e-4176-bfb8-3be9025f9f27", "type": "job-post", "date": "2026-07-16", "quote": "a large GPU fleet spread across several cloud providers" } ] }, { "tool": "TensorRT-LLM", "vendor": "NVIDIA", "layer": "inference", "usage": "Served LLMs on GPU pods before the in-house engine took over.", "status": "historical", "confidence": "high", "last_verified": "2026-07-26", "added": "2026-07-26", "evidence": [ { "url": "https://developer.nvidia.com/blog/spotlight-perplexity-ai-serves-400-million-search-queries-a-month-using-nvidia-inference-stack", "type": "case-study", "date": "2024-12-05", "quote": "Triton Inference Server is a critical component of Perplexity's deployment architecture." }, { "url": "https://jobs.ashbyhq.com/perplexity/8a976851-9bef-4b07-8d36-567fa9540aef", "type": "job-post", "date": "2026-04-13", "quote": "Our stack is Rust, Python, CUDA, and CuTe DSL" } ] }, { "tool": "Cerebras Inference", "vendor": "Cerebras", "layer": "inference", "usage": "Wafer-scale inference serving the Sonar model for near-instant answers.", "status": "current", "confidence": "high", "last_verified": "2026-07-26", "added": "2026-07-26", "evidence": [ { "url": "https://www.cerebras.ai/press-release/cerebras-powers-perplexity-sonar-with-industrys-fastest-ai-inference", "type": "press", "date": "2025-02-11", "quote": "1,200 tokens per second, delivering near-instant answer generation" } ] }, { "tool": "TransferEngine (pplx-garden)", "vendor": "Perplexity", "layer": "inference", "usage": "Open-sourced RDMA library powering disaggregated serving of large MoE models.", "status": "current", "confidence": "high", "last_verified": "2026-07-26", "added": "2026-07-26", "evidence": [ { "url": "https://arxiv.org/abs/2510.27656", "type": "docs", "date": "2025-10-31", "quote": "KvCache transfer for disaggregated inference with dynamic scaling" }, { "url": "https://github.com/perplexityai/pplx-garden", "type": "github", "date": "2025-11-04", "quote": "Perplexity open source garden for inference technology" } ] }, { "tool": "Model Context Protocol", "vendor": "Anthropic", "layer": "orchestration", "usage": "Agent-integration standard with an official server exposing Sonar and Search.", "status": "current", "confidence": "high", "last_verified": "2026-07-26", "added": "2026-07-26", "evidence": [ { "url": "https://github.com/perplexityai/modelcontextprotocol", "type": "github", "date": "2025-03-10", "quote": "The official MCP server implementation for the Perplexity API Platform" } ] }, { "tool": "Vespa", "vendor": "Vespa.ai", "layer": "rag-vector", "usage": "Retrieval and ranking platform for the in-house web index.", "status": "current", "confidence": "high", "last_verified": "2026-07-26", "added": "2026-07-26", "evidence": [ { "url": "https://blog.vespa.ai/perplexity-builds-ai-search-at-scale-on-vespa-ai/", "type": "case-study", "date": "2025-04-15", "quote": "Perplexity has been able to create their own search engine that delivers uncompromising quality" }, { "url": "https://vespa.ai/perplexity-partners-with-vespa-ai-to-bring-its-search-function-in-house/", "type": "press", "date": "2025-04-15", "quote": "Perplexity has used Vespa.ai's managed platform to efficiently scale its RAG architecture" } ] }, { "tool": "In-house web index (PerplexityBot)", "vendor": "Perplexity", "layer": "rag-vector", "usage": "Proprietary crawler and index feeding hybrid lexical and semantic retrieval.", "status": "current", "practice": true, "confidence": "high", "last_verified": "2026-07-26", "added": "2026-07-26", "evidence": [ { "url": "https://research.perplexity.ai/articles/architecting-and-evaluating-an-ai-first-search-api", "type": "eng-blog", "date": "2026-07-17", "quote": "Our search index tracks over 200 billion unique URLs, with capacity to track many hundreds of billions more." } ] }, { "tool": "In-house evals (LLM-as-judge)", "vendor": "Perplexity", "layer": "evals-observability", "usage": "Grades search quality with LLM-as-judge evaluations over public benchmarks.", "status": "current", "practice": true, "confidence": "high", "last_verified": "2026-07-26", "added": "2026-07-26", "evidence": [ { "url": "https://research.perplexity.ai/articles/architecting-and-evaluating-an-ai-first-search-api", "type": "eng-blog", "date": "2026-07-17", "quote": "We grade all benchmarks using the same prompted classifier methodology used in the original work" }, { "url": "https://jobs.ashbyhq.com/perplexity/4615ca06-bea7-47e3-9e57-f5cee52b75e6", "type": "job-post", "date": "2026-06-29", "quote": "you will build specialized evals to improve answer quality across Perplexity" } ] }, { "tool": "Amazon SageMaker HyperPod", "vendor": "AWS", "layer": "training", "usage": "Early distributed training platform, since replaced by a self-managed GPU fleet.", "status": "historical", "confidence": "medium", "last_verified": "2026-07-26", "added": "2026-07-26", "evidence": [ { "url": "https://aws.amazon.com/solutions/case-studies/perplexity-case-study/", "type": "case-study", "date": "2024-03", "quote": "Amazon SageMaker HyperPod's built-in data and model parallel libraries helped us optimize training time on GPUs" }, { "url": "https://www.youtube.com/watch?v=0JoodauiFds", "type": "talk", "date": "2024-03-27", "quote": "Perplexity Accelerates Foundation Model Training by 40% with Amazon SageMaker HyperPod" } ] }, { "tool": "NeMo", "vendor": "NVIDIA", "layer": "training", "usage": "Framework behind the post-training run that produced R1-1776.", "status": "current", "confidence": "medium", "last_verified": "2026-07-26", "added": "2026-07-26", "evidence": [ { "url": "https://hyperight.com/perplexity-r1-1776-matches-deepseek-r1-performance-without-the-censorship/", "type": "press", "date": "2025-02-24", "quote": "They used NVIDIA's NeMo 2.0 framework to fine-tune the model" } ] }, { "tool": "In-house data curation", "vendor": "Perplexity", "layer": "data-labeling", "usage": "Curates multilingual post-training datasets with in-house experts.", "status": "current", "practice": true, "confidence": "medium", "last_verified": "2026-07-26", "added": "2026-07-26", "evidence": [ { "url": "https://hyperight.com/perplexity-r1-1776-matches-deepseek-r1-performance-without-the-censorship/", "type": "press", "date": "2025-02-24", "quote": "they built a dataset of 40,000 multilingual prompts to retrain the model" } ] }, { "tool": "BrowseSafe", "vendor": "Perplexity", "layer": "guardrails", "usage": "Fine-tuned classifier detecting prompt-injection attacks in the Comet browser.", "status": "current", "confidence": "high", "last_verified": "2026-07-26", "added": "2026-07-26", "evidence": [ { "url": "https://x.com/AravSrinivas/status/1996028650727333928", "type": "social", "date": "2025-12-03", "quote": "fine-tuned a version of Qwen3-30B that can scan raw HTML and detect prompt injection attacks" }, { "url": "https://mpost.io/perplexity-ai-open-sources-browsesafe-to-combat-prompt-injection-in-ai-browsing/", "type": "press", "date": "2025-12-04", "quote": "BrowseSafe is designed to analyze complete webpages quickly without affecting browser performance." } ] }, { "tool": "Trail of Bits security audit", "vendor": "Trail of Bits", "layer": "guardrails", "usage": "External security firm stress-tests Comet's agentic browsing against prompt injection.", "status": "current", "practice": true, "confidence": "high", "last_verified": "2026-07-26", "added": "2026-07-26", "evidence": [ { "url": "https://blog.trailofbits.com/2026/02/20/using-threat-modeling-and-prompt-injection-to-audit-comet/", "type": "eng-blog", "date": "2026-02-20", "quote": "Before launching their Comet browser, Perplexity hired us to test the security of their AI-powered browsing features." } ] }, { "tool": "Cursor", "vendor": "Anysphere", "layer": "ai-coding", "usage": "Mandatory AI coding tool used across engineering and beyond.", "status": "current", "confidence": "medium", "last_verified": "2026-07-26", "added": "2026-07-26", "evidence": [ { "url": "https://www.aol.com/perplexitys-engineers-2-ai-coding-061356405.html", "type": "press", "date": "2025-07-14", "quote": "They upload my screenshot to Cursor and then ask it to write a change to the Swift UI file" }, { "url": "https://www.entrepreneur.com/business-news/perplexity-ceo-ai-coding-tools-transformed-the-way-we-work/494663", "type": "press", "date": "2025-07-15", "quote": "made it compulsory" } ] }, { "tool": "Internal AI data agents", "vendor": "Perplexity", "layer": "internal-ai", "usage": "Internal agents automate end-to-end analytics workflows for the data org.", "status": "current", "practice": true, "confidence": "medium", "last_verified": "2026-07-26", "added": "2026-07-26", "evidence": [ { "url": "https://jobs.ashbyhq.com/perplexity/a14c4626-ed07-4151-a923-3a710fb66de2", "type": "job-post", "date": "2026-07-23", "quote": "You'll join a data team that's already using AI across its work." } ] } ] }, { "slug": "physical-intelligence", "name": "Physical Intelligence", "website": "https://www.physicalintelligence.company", "what": "Robotics foundation model company.", "profile": "Physical Intelligence runs real-time remote inference for robotic control on Modal, allocating data-center-class GPUs per deployment for GPU-intensive experiments.", "facts": [ { "tool": "Modal", "vendor": "Modal", "layer": "inference", "usage": "Runs real-time remote inference for robotic control.", "status": "current", "confidence": "high", "last_verified": "2026-07-27", "added": "2026-07-27", "evidence": [ { "url": "https://modal.com/blog/physical-intelligence-runs-real-time-remote-inference-for-robotic-control-on-modal", "type": "case-study", "date": "2026-04-08", "quote": "On Modal, PI can allocate larger, data-center-class GPUs per deployment and run GPU-intensive experiments immediately." } ] } ] }, { "slug": "podium", "name": "Podium", "website": "https://www.podium.com", "what": "AI-powered lead conversion and communication platform for local businesses.", "profile": "Podium's AI employee started on the LangChain framework, added LangSmith for LLM testing and observability, and integrated LangGraph to reduce agent orchestration complexity and make conversations more controllable.", "facts": [ { "tool": "LangSmith", "vendor": "LangChain", "layer": "evals-observability", "usage": "Tests and monitors AI employee performance with traces and datasets.", "status": "current", "confidence": "high", "last_verified": "2026-07-27", "added": "2026-07-27", "evidence": [ { "url": "https://blog.langchain.dev/customers-podium/", "type": "case-study", "date": "2024-08-15", "quote": "turned to LangSmith for LLM testing and observability." } ] }, { "tool": "LangChain", "vendor": "LangChain", "layer": "orchestration", "usage": "Underpins single-turn interactions in its AI employee.", "status": "current", "confidence": "medium", "last_verified": "2026-07-27", "added": "2026-07-27", "evidence": [ { "url": "https://blog.langchain.dev/customers-podium/", "type": "case-study", "date": "2024-08-15", "quote": "Initially, Podium used the LangChain framework for single-turn interactions." } ] }, { "tool": "LangGraph", "vendor": "LangChain", "layer": "orchestration", "usage": "Reduces orchestration complexity and increases conversation controllability.", "status": "current", "confidence": "high", "last_verified": "2026-07-27", "added": "2026-07-27", "evidence": [ { "url": "https://blog.langchain.dev/customers-podium/", "type": "case-study", "date": "2024-08-15", "quote": "integrating LangGraph into its workflow, reducing complexity in their agent orchestration" } ] } ] }, { "slug": "portola", "name": "Portola", "website": "https://www.tolans.com", "what": "Maker of Tolan, an AI alien companion app.", "profile": "Portola curates datasets in Braintrust tagged to specific conversation issues and iterates on prompts in playgrounds to improve its AI companions' conversation quality.", "facts": [ { "tool": "Braintrust", "vendor": "Braintrust", "layer": "evals-observability", "usage": "Curates datasets and iterates on prompts for conversation quality.", "status": "current", "confidence": "high", "last_verified": "2026-07-27", "added": "2026-07-27", "evidence": [ { "url": "https://www.braintrust.dev/customers/portola", "type": "case-study", "date": "2026-07", "quote": "she creates a dataset in Braintrust tagged with the specific issue." } ] } ] }, { "slug": "predori", "name": "predori", "website": "https://predori.com", "what": "AI patent research and monitoring software.", "profile": "predori has run its patent intelligence platform on Weaviate Cloud since 2021, powering vector retrieval and RAG.", "facts": [ { "tool": "Weaviate", "vendor": "Weaviate", "layer": "rag-vector", "usage": "Powers patent intelligence vector retrieval and RAG.", "status": "current", "confidence": "high", "last_verified": "2026-07-27", "added": "2026-07-27", "evidence": [ { "url": "https://weaviate.io/case-studies/predori", "type": "case-study", "date": "2026-07", "quote": "chose Weaviate Cloud in 2021." } ] } ] }, { "slug": "pylon", "name": "Pylon", "website": "https://usepylon.com", "what": "Agentic B2B customer support platform.", "profile": "Pylon turned eval discipline into a merge rule — every AI prompt checked into the codebase needs a Braintrust playground ID or the CI test fails, and the curated dataset behind that playground travels with the prompt. Eval sets are distilled from the rare failure modes in production traffic, scorers are spun up with Loop, and a Claude Code skill reconstructs incidents through the Braintrust MCP.", "facts": [ { "tool": "Braintrust", "vendor": "Braintrust", "layer": "evals-observability", "usage": "Tests every prompt in CI and observes production traffic.", "status": "current", "confidence": "high", "last_verified": "2026-08-22", "added": "2026-08-22", "evidence": [ { "url": "https://www.braintrust.dev/customers/pylon", "type": "case-study", "date": "2026-08", "quote": "requires using Braintrust to test every prompt as part of the CI pipeline." } ] }, { "tool": "Loop", "vendor": "Braintrust", "layer": "evals-observability", "usage": "Spins up new LLM-as-a-judge scorers for bulk quality checks.", "status": "current", "confidence": "high", "last_verified": "2026-08-22", "added": "2026-08-22", "evidence": [ { "url": "https://www.braintrust.dev/customers/pylon", "type": "case-study", "date": "2026-08", "quote": "uses Loop to spin up new scorers quickly." } ] } ] }, { "slug": "quora", "name": "Quora", "website": "https://www.quora.com", "what": "Question-and-answer platform and operator of Poe, a multi-model AI chat app.", "profile": "Quora securely executes LLM-generated code in Poe, its AI chatbot platform, using Modal Sandboxes.", "facts": [ { "tool": "Modal Sandboxes", "vendor": "Modal", "layer": "guardrails", "usage": "Securely executes LLM-generated code in Poe.", "status": "current", "confidence": "high", "last_verified": "2026-07-27", "added": "2026-07-27", "evidence": [ { "url": "https://modal.com/blog/quora-case-study", "type": "case-study", "date": "2025-06-30", "quote": "Quora uses Modal Sandboxes to securely execute LLM-generated code in Poe, their AI chatbot platform." } ] } ] }, { "slug": "rainn", "name": "RAINN", "website": "https://rainn.org", "what": "Largest US anti-sexual-violence organization, operating the National Sexual Assault Hotline.", "profile": "RAINN runs much of its operations on Claude and uses Claude Code to build infrastructure for crisis support integrations.", "facts": [ { "tool": "Claude Code", "vendor": "Anthropic", "layer": "ai-coding", "usage": "Builds infrastructure for crisis support integrations.", "status": "current", "confidence": "high", "last_verified": "2026-07-27", "added": "2026-07-27", "evidence": [ { "url": "https://claude.com/customers/rainn", "type": "case-study", "date": "2026-07", "quote": "uses Claude Code to build infrastructure for crisis support" } ] }, { "tool": "Claude", "vendor": "Anthropic", "layer": "internal-ai", "usage": "Runs much of the organization's day-to-day operations.", "status": "current", "confidence": "high", "last_verified": "2026-07-27", "added": "2026-07-27", "evidence": [ { "url": "https://claude.com/customers/rainn", "type": "case-study", "date": "2026-07", "quote": "run much of RAINN’s operations on Claude." } ] } ] }, { "slug": "rakuten-group", "name": "Rakuten Group", "website": "https://global.rakuten.com/corp/", "what": "Japanese internet conglomerate spanning e-commerce, fintech, and digital content.", "profile": "Rakuten builds AI products for business clients and employees on LangChain — including an OpenGPTs-based employee assistant — and uses LangSmith to monitor agent performance and distribute prompts across teams.", "facts": [ { "tool": "LangChain", "vendor": "LangChain", "layer": "orchestration", "usage": "Builds AI solutions and employee chatbots for clients and staff.", "status": "current", "confidence": "high", "last_verified": "2026-07-27", "added": "2026-07-27", "evidence": [ { "url": "https://blog.langchain.dev/customers-rakuten/", "type": "case-study", "date": "2024-02-14", "quote": "Rakuten recently leveraged LangChain’s OpenGPTs package to deliver an employee empowerment experience" } ] }, { "tool": "LangSmith", "vendor": "LangChain", "layer": "evals-observability", "usage": "Monitors agent performance and distributes prompts across teams.", "status": "current", "confidence": "high", "last_verified": "2026-07-27", "added": "2026-07-27", "evidence": [ { "url": "https://blog.langchain.dev/customers-rakuten/", "type": "case-study", "date": "2024-02-14", "quote": "LangSmith allows us to get things done scientifically." } ] } ] }, { "slug": "ramp", "name": "Ramp", "website": "https://ramp.com", "what": "Finance automation platform for corporate cards, expenses, and payments.", "profile": "Ramp built Ramp Inspect, a full-context background coding agent, on Modal — spinning up full development environments in seconds with Modal Sandboxes — and uses the Vercel AI SDK to build semantic search across internal assets. Engineering runs on Claude Code, including an On-call Assistant that root-causes every incident, and the team has put Claude Fable on large refactors of its Python monolith.", "facts": [ { "tool": "Modal", "vendor": "Modal", "layer": "inference", "usage": "Powers the Ramp Inspect background coding agent.", "status": "current", "confidence": "high", "last_verified": "2026-07-27", "added": "2026-07-27", "evidence": [ { "url": "https://modal.com/blog/how-ramp-built-a-full-context-background-coding-agent-on-modal", "type": "case-study", "date": "2026-02-19", "quote": "Ramp uses Modal to power Ramp Inspect" }, { "url": "https://modal.com/blog/how-ramp-built-a-full-context-background-coding-agent-on-modal", "type": "case-study", "date": "2026-02-19", "quote": "Leveraging Modal Sandboxes, Ramp spins up full development environments in seconds," } ] }, { "tool": "Vercel AI SDK", "vendor": "Vercel", "layer": "orchestration", "usage": "Builds semantic search across scattered internal assets.", "status": "current", "confidence": "high", "last_verified": "2026-07-27", "added": "2026-07-27", "evidence": [ { "url": "https://vercel.com/customers/how-ramp-kept-100-uptime-through-100x-traffic-surges-on-vercel", "type": "case-study", "date": "2026-02-17", "quote": "The team is now using the Vercel AI SDK to build semantic search across disparate internal assets." } ] }, { "tool": "Claude Code", "vendor": "Anthropic", "layer": "ai-coding", "usage": "Runs the On-call Assistant and engineers' local development work.", "status": "current", "confidence": "high", "last_verified": "2026-08-22", "added": "2026-08-22", "evidence": [ { "url": "https://claude.com/customers/ramp-boris-office-hours", "type": "case-study", "date": "2026-08", "quote": "On-call Assistant, which has always run on Claude Code" } ] }, { "tool": "Claude Fable", "vendor": "Anthropic", "layer": "models", "usage": "Applied to large refactors in the Python monolith, with code merged.", "status": "current", "confidence": "high", "last_verified": "2026-08-22", "added": "2026-08-22", "evidence": [ { "url": "https://claude.com/customers/ramp-boris-office-hours", "type": "case-study", "date": "2026-08", "quote": "I told Fable to fix all our import cycles." } ] } ] }, { "slug": "reducto", "name": "Reducto", "website": "https://reducto.ai", "what": "Document ingestion and parsing APIs for LLM pipelines.", "profile": "Reducto scales GPU infrastructure for multi-model document processing on Modal, expanding its deployment of large language and vision-language models.", "facts": [ { "tool": "Modal", "vendor": "Modal", "layer": "inference", "usage": "Scales GPU inference for multi-model document processing.", "status": "current", "confidence": "high", "last_verified": "2026-07-27", "added": "2026-07-27", "evidence": [ { "url": "https://modal.com/blog/reducto-case-study", "type": "case-study", "date": "2025-11-19", "quote": "Reducto continues to expand its use of Modal with the deployment of new large language and vision-language models." } ] } ] }, { "slug": "retool", "name": "Retool", "website": "https://retool.com", "what": "Enterprise platform for building internal tools and apps.", "profile": "Retool validates AI classifier accuracy through iterative testing in Braintrust, which the team credits as central to executing its AI roadmap.", "facts": [ { "tool": "Braintrust", "vendor": "Braintrust", "layer": "evals-observability", "usage": "Evaluates AI classifier accuracy through iterative testing.", "status": "current", "confidence": "high", "last_verified": "2026-07-27", "added": "2026-07-27", "evidence": [ { "url": "https://www.braintrust.dev/customers/retool", "type": "case-study", "date": "2026-07", "quote": "Braintrust has been the lifeblood of our ability to execute against our roadmap" } ] } ] }, { "slug": "runware", "name": "Runware", "website": "https://runware.ai", "what": "Generative media inference API platform.", "profile": "Runware taps Together for immediate access to NVIDIA H100, H200, and B200 GPUs, giving it on-demand capacity for rapid model deployment and scaling.", "facts": [ { "tool": "Together GPU Clusters", "vendor": "Together AI", "layer": "inference", "usage": "Provides on-demand GPU capacity for rapid model deployment.", "status": "current", "confidence": "high", "last_verified": "2026-07-27", "added": "2026-07-27", "evidence": [ { "url": "https://www.together.ai/customers/runware", "type": "case-study", "date": "2026-07", "quote": "Together provides immediate access to NVIDIA H100s, H200s, and B200s without long-term commitments." } ] } ] }, { "slug": "scaled-cognition", "name": "Scaled Cognition", "website": "https://www.scaledcognition.com", "what": "AI lab training agentic models for customer-facing tasks.", "profile": "Scaled Cognition runs custom model training on Together AI's GPU Clusters with direct access to H200 and B200 hardware.", "facts": [ { "tool": "Together GPU Clusters", "vendor": "Together AI", "layer": "training", "usage": "Provides bare-metal GPU infrastructure for custom model training.", "status": "current", "confidence": "high", "last_verified": "2026-07-27", "added": "2026-07-27", "evidence": [ { "url": "https://www.together.ai/customers/scaled-cognition", "type": "case-study", "date": "2026-07", "quote": "How Scaled Cognition Trains APT-1 on Together AI GPU Clusters" } ] } ] }, { "slug": "sensay", "name": "Sensay", "website": "https://sensay.io", "what": "Platform for AI replicas and digital twins.", "profile": "Sensay streamlines building and maintaining its AI features with the Vercel AI SDK.", "facts": [ { "tool": "Vercel AI SDK", "vendor": "Vercel", "layer": "orchestration", "usage": "Streamlines building and maintaining AI features.", "status": "current", "confidence": "high", "last_verified": "2026-07-27", "added": "2026-07-27", "evidence": [ { "url": "https://vercel.com/customers/how-sensay-went-from-zero-to-product-in-six-weeks", "type": "case-study", "date": "2026-01-27", "quote": "use the Vercel AI SDK to streamline" } ] } ] }, { "slug": "sentient", "name": "Sentient", "website": "https://sentient.xyz", "what": "Open-source AGI research organization building decentralized AI.", "profile": "Sentient runs its multi-agent chat and search products on Fireworks AI for high-concurrency inference.", "facts": [ { "tool": "Fireworks AI", "vendor": "Fireworks AI", "layer": "inference", "usage": "Powers multi-agent chat and search with high-concurrency inference.", "status": "current", "confidence": "medium", "last_verified": "2026-07-27", "added": "2026-07-27", "evidence": [ { "url": "https://fireworks.ai/blog/Story-Sentient", "type": "case-study", "date": "2025-07-17", "quote": "It was running on Fireworks." } ] } ] }, { "slug": "servicenow", "name": "ServiceNow", "website": "https://www.servicenow.com", "what": "Enterprise platform for IT, employee, and customer workflows.", "profile": "ServiceNow orchestrates multi-agent customer-success workflows with LangGraph — integrated with its knowledge graph and MCP — and runs a tailored evaluation framework in LangSmith to monitor live agent performance.", "facts": [ { "tool": "LangSmith", "vendor": "LangChain", "layer": "evals-observability", "usage": "Runs a tailored evaluation framework over its multi-agent system.", "status": "current", "confidence": "high", "last_verified": "2026-07-27", "added": "2026-07-27", "evidence": [ { "url": "https://blog.langchain.com/customers-servicenow/", "type": "case-study", "date": "2025-11-17", "quote": "ServiceNow implemented a sophisticated evaluation framework in LangSmith tailored to their multi-agent system." } ] }, { "tool": "LangGraph", "vendor": "LangChain", "layer": "orchestration", "usage": "Orchestrates multi-agent workflows with knowledge-graph and MCP integration.", "status": "current", "confidence": "high", "last_verified": "2026-07-27", "added": "2026-07-27", "evidence": [ { "url": "https://blog.langchain.com/customers-servicenow/", "type": "case-study", "date": "2025-11-17", "quote": "ServiceNow has integrated their knowledge graph and Model Context Protocol (MCP) with LangGraph" } ] } ] }, { "slug": "shopify", "name": "Shopify", "website": "https://www.shopify.com", "what": "E-commerce platform powering millions of merchant storefronts and an AI merchant assistant (Sidekick).", "profile": "Shopify runs one of the most public \"AI-first\" programs in tech: Tobi Lütke's April 2025 memo made reflexive AI usage a baseline expectation, and the company ships merchant-facing AI (Sidekick on Claude Sonnet via Vertex AI, the Shopify Magic suite, SimGym simulated shoppers) on top of a deliberately multi-provider model strategy. Their build-vs-buy philosophy is distinctive: buy frontier models and coding tools at scale (Cursor, Copilot, Claude Code), but build and open-source the orchestration and internal layers themselves (Roast, MCP-UI, a LibreChat fork for chat.shopify.io) and fine-tune open-source VLMs where API costs are prohibitive, with in-house LLM-judge evals and GRPO post-training for Sidekick.", "facts": [ { "tool": "Claude Sonnet", "vendor": "Anthropic", "layer": "models", "usage": "Primary model behind Sidekick, the AI commerce assistant for merchants.", "status": "current", "confidence": "high", "last_verified": "2026-07-26", "added": "2026-07-26", "evidence": [ { "url": "https://cloud.google.com/blog/products/ai-machine-learning/expanding-vertex-ai-with-claude-opus-4-6", "type": "case-study", "date": "2026-02-05", "quote": "The combination of Claude and Vertex AI helps us empower millions of merchants with our AI-enabled commerce assistant, Sidekick" }, { "url": "https://help.shopify.com/en/manual/privacy-and-security/privacy/subprocessors", "type": "subprocessor", "date": "2026-07", "quote": "Conversational artificial intelligence services" } ] }, { "tool": "OpenAI API", "vendor": "OpenAI", "layer": "models", "usage": "Additional model provider behind Sidekick's multi-provider routing.", "status": "current", "confidence": "high", "last_verified": "2026-07-26", "added": "2026-07-26", "evidence": [ { "url": "https://help.shopify.com/en/manual/privacy-and-security/privacy/subprocessors", "type": "subprocessor", "date": "2026-07", "quote": "Personal data if input to or included in a response by Shopify Sidekick" } ] }, { "tool": "Qwen2VL 7B (fine-tuned)", "vendor": "Alibaba (open source)", "layer": "models", "usage": "Fine-tuned vision-language model powering the product-understanding pipeline.", "status": "current", "confidence": "high", "last_verified": "2026-07-26", "added": "2026-07-26", "evidence": [ { "url": "https://shopify.engineering/leveraging-multimodal-llms", "type": "eng-blog", "date": "2025-07-16", "quote": "We have deployed three successive open-source models: LlaVA 1.5 7B, LLaMA 3.2 11B, and currently Qwen2VL 7B." } ] }, { "tool": "Shopify Magic", "vendor": "Shopify (in-house)", "layer": "models", "usage": "Free AI feature suite built on multiple leading model providers.", "status": "current", "confidence": "high", "last_verified": "2026-07-26", "added": "2026-07-26", "evidence": [ { "url": "https://help.shopify.com/en/manual/shopify-admin/productivity-tools/shopify-magic", "type": "docs", "date": "2026-07", "quote": "built on a combination of Shopify's proprietary data and the world's leading large language models" } ] }, { "tool": "Vertex AI", "vendor": "Google Cloud", "layer": "inference", "usage": "Serves Claude models for Sidekick at merchant scale.", "status": "current", "confidence": "high", "last_verified": "2026-07-26", "added": "2026-07-26", "evidence": [ { "url": "https://cloud.google.com/blog/products/ai-machine-learning/expanding-vertex-ai-with-claude-opus-4-6", "type": "case-study", "date": "2026-02-05", "quote": "The combination of Claude and Vertex AI helps us empower millions of merchants with our AI-enabled commerce assistant, Sidekick" } ] }, { "tool": "Triton Inference Server", "vendor": "NVIDIA", "layer": "inference", "usage": "Self-hosts fine-tuned vision models across the GPU fleet.", "status": "current", "confidence": "high", "last_verified": "2026-07-26", "added": "2026-07-26", "evidence": [ { "url": "https://shopify.engineering/leveraging-multimodal-llms", "type": "eng-blog", "date": "2025-07-16", "quote": "Orchestrates model serving across our GPU fleet, handling request preprocessing, batching, and routing" }, { "url": "https://shopify.engineering/leveraging-multimodal-llms", "type": "eng-blog", "date": "2025-07-16", "quote": "40 million LLM calls daily, representing about 16 billion tokens inferred per day" } ] }, { "tool": "GroqCloud", "vendor": "Groq", "layer": "inference", "usage": "AI services provider whose exact workload is not publicly detailed.", "status": "current", "confidence": "medium", "last_verified": "2026-07-26", "added": "2026-07-26", "evidence": [ { "url": "https://help.shopify.com/en/manual/privacy-and-security/privacy/subprocessors", "type": "subprocessor", "date": "2026-07", "quote": "Artificial intelligence services" } ] }, { "tool": "Roast", "vendor": "Shopify (open source)", "layer": "orchestration", "usage": "Home-grown Ruby DSL for structured AI workflows.", "status": "current", "confidence": "high", "last_verified": "2026-07-26", "added": "2026-07-26", "evidence": [ { "url": "https://shopify.engineering/introducing-roast", "type": "eng-blog", "date": "2025-06-18", "quote": "Non-determinism is the enemy of reliability." }, { "url": "https://github.com/Shopify/roast", "type": "github", "date": "2026-07", "quote": "Build complex AI-powered automation with simple, declarative Ruby syntax." } ] }, { "tool": "Model Context Protocol (MCP)", "vendor": "Anthropic (open standard)", "layer": "orchestration", "usage": "Exposes internal company data and storefronts to agents over MCP.", "status": "current", "confidence": "high", "last_verified": "2026-07-26", "added": "2026-07-26", "evidence": [ { "url": "https://shopify.dev/docs/apps/build/storefront-mcp", "type": "docs", "date": "2026-07", "quote": "Connect any AI assistant to real-time commerce data from Shopify stores" }, { "url": "https://www.firstround.com/ai/shopify", "type": "press", "date": "2025-07-15", "quote": "We make every single piece of data inside the company available" } ] }, { "tool": "Google Cloud Dataflow", "vendor": "Google Cloud", "layer": "rag-vector", "usage": "Streams real-time embedding pipelines feeding storefront semantic search.", "status": "current", "confidence": "high", "last_verified": "2026-07-26", "added": "2026-07-26", "evidence": [ { "url": "https://shopify.engineering/how-shopify-improved-consumer-search-intent-with-real-time-ml", "type": "eng-blog", "date": "2024-10-18", "quote": "we use Google Cloud's streaming analytics service Dataflow to power these pipelines" } ] }, { "tool": "In-house LLM-judge eval platform (GTX + merchant simulator)", "vendor": "Shopify (in-house)", "layer": "evals-observability", "usage": "Evaluates Sidekick with calibrated LLM judges and a merchant simulator.", "status": "current", "practice": true, "confidence": "high", "last_verified": "2026-07-26", "added": "2026-07-26", "evidence": [ { "url": "https://shopify.engineering/building-production-ready-agentic-systems", "type": "eng-blog", "date": "2025-08-26", "quote": "we built an LLM-powered merchant simulator that captures the 'essence' or goals of real conversations" }, { "url": "https://shopify.engineering/building-production-ready-agentic-systems", "type": "eng-blog", "date": "2025-08-26", "quote": "We moved away from carefully curated \"golden\" datasets toward Ground Truth Sets (GTX)" } ] }, { "tool": "GRPO reinforcement-learning fine-tuning", "vendor": "Shopify (in-house)", "layer": "training", "usage": "Post-trains Sidekick's fine-tuned models with LLM judges as reward signals.", "status": "current", "practice": true, "confidence": "high", "last_verified": "2026-07-26", "added": "2026-07-26", "evidence": [ { "url": "https://shopify.engineering/building-production-ready-agentic-systems", "type": "eng-blog", "date": "2025-08-26", "quote": "we implemented Group Relative Policy Optimization (GRPO), a reinforcement learning approach that uses our LLM judges as reward signals" }, { "url": "https://shopify.engineering/sidekick-curation", "type": "eng-blog", "date": "2026-06-15", "quote": "Fine-tuning an open-source model got us there for the common cases." } ] }, { "tool": "Toloka", "vendor": "Toloka", "layer": "data-labeling", "usage": "Supplies the human-annotated seed dataset calibrating automated training-data curation.", "status": "current", "confidence": "high", "last_verified": "2026-07-26", "added": "2026-07-26", "evidence": [ { "url": "https://shopify.engineering/sidekick-curation", "type": "eng-blog", "date": "2026-06-15", "quote": "We initially partnered with the Toloka team to produce a balanced dataset" }, { "url": "https://shopify.engineering/sidekick-curation", "type": "eng-blog", "date": "2026-06-15", "quote": "If four independent models can't agree on a label, a human should make the call" } ] }, { "tool": "Refusal-behavior training pipeline", "vendor": "Shopify (in-house)", "layer": "guardrails", "usage": "Trains Sidekick to refuse impossible or out-of-scope requests.", "status": "current", "practice": true, "confidence": "high", "last_verified": "2026-07-26", "added": "2026-07-26", "evidence": [ { "url": "https://shopify.engineering/sidekick-curation", "type": "eng-blog", "date": "2026-06-15", "quote": "The model had never learned to say no." } ] }, { "tool": "Cursor", "vendor": "Anysphere", "layer": "ai-coding", "usage": "Standard AI code editor across engineering.", "status": "current", "confidence": "high", "last_verified": "2026-07-26", "added": "2026-07-26", "evidence": [ { "url": "https://www.firstround.com/ai/shopify", "type": "press", "date": "2025-07-15", "quote": "I ordered 1,500 Cursor licenses last year and quickly had to procure another 1,500." }, { "url": "https://x.com/tobi/status/1909251946235437514", "type": "social", "date": "2025-04-07", "quote": "Developers have proxy, Copilot, Cursor, Claude code" } ] }, { "tool": "GitHub Copilot", "vendor": "GitHub (Microsoft)", "layer": "ai-coding", "usage": "Longest-standing AI coding tool in the sanctioned developer toolkit.", "status": "current", "confidence": "high", "last_verified": "2026-07-26", "added": "2026-07-26", "evidence": [ { "url": "https://www.firstround.com/ai/shopify", "type": "press", "date": "2025-07-15", "quote": "Pretty quickly, our adoption increased to 80% and GitHub started asking us how we did that" }, { "url": "https://x.com/tobi/status/1909251946235437514", "type": "social", "date": "2025-04-07", "quote": "Developers have proxy, Copilot, Cursor, Claude code" } ] }, { "tool": "Claude Code", "vendor": "Anthropic", "layer": "ai-coding", "usage": "Sanctioned agentic coding tool, also embedded as Roast's coding agent.", "status": "current", "confidence": "high", "last_verified": "2026-07-26", "added": "2026-07-26", "evidence": [ { "url": "https://techstartups.com/2025/04/07/shopify-ceo-tobi-lutke-confirms-leaked-internal-memo-on-social-media-about-hiring-ai-before-humans/", "type": "press", "date": "2025-04-07", "quote": "There is chat.shopify.io, which we had for years now. Developers have proxy, Copilot, Cursor, Claude code" }, { "url": "https://claude.com/customers/shopify", "type": "case-study", "date": "2026-07", "quote": "Claude Code has transformed how we build internal tools at Shopify" } ] }, { "tool": "LibreChat (chat.shopify.io)", "vendor": "LibreChat (open source)", "layer": "internal-ai", "usage": "Runs the company-wide internal AI chat on an internal fork.", "status": "current", "confidence": "high", "last_verified": "2026-07-26", "added": "2026-07-26", "evidence": [ { "url": "https://x.com/tobi/status/1932846291794510241", "type": "social", "date": "2025-06-11", "quote": "Shopify runs an internal fork of librechat, and we merge most everything back." }, { "url": "https://www.firstround.com/ai/shopify", "type": "press", "date": "2025-07-15", "quote": "Built using LibreChat (to which Shopify is a core contributor)" } ] }, { "tool": "Reflexive AI usage mandate", "vendor": "Shopify (internal policy)", "layer": "internal-ai", "usage": "Company-wide policy making reflexive AI use mandatory in daily work.", "status": "current", "practice": true, "confidence": "high", "last_verified": "2026-07-26", "added": "2026-07-26", "evidence": [ { "url": "https://x.com/tobi/status/1909251946235437514", "type": "social", "date": "2025-04-07", "quote": "Reflexive AI usage is now a baseline expectation at Shopify" }, { "url": "https://techstartups.com/2025/04/07/shopify-ceo-tobi-lutke-confirms-leaked-internal-memo-on-social-media-about-hiring-ai-before-humans/", "type": "press", "date": "2025-04-07", "quote": "Before asking for more Headcount and resources, teams must demonstrate why they cannot get what they want done using AI" } ] } ] }, { "slug": "slack", "name": "Slack", "website": "https://slack.com", "what": "Workplace messaging and collaboration platform, part of Salesforce.", "profile": "Slack builds its AI features — summarization, search, and organizational knowledge access — on Anthropic's Claude, selected for its strength in nuanced conversation analysis. Its engineering team also uses Claude Code to fix bugs and move faster.", "facts": [ { "tool": "Claude", "vendor": "Anthropic", "layer": "models", "usage": "Powers Slack AI summarization, search, and knowledge features.", "status": "current", "confidence": "high", "last_verified": "2026-07-27", "added": "2026-07-27", "evidence": [ { "url": "https://www.claude.com/customers/slack", "type": "case-study", "date": "2026-07", "quote": "Slack selected Claude for its exceptional capability in human-like understanding and nuanced conversation analysis." } ] }, { "tool": "Claude Code", "vendor": "Anthropic", "layer": "ai-coding", "usage": "Fixes bugs and accelerates engineering workflows.", "status": "current", "confidence": "high", "last_verified": "2026-07-27", "added": "2026-07-27", "evidence": [ { "url": "https://www.claude.com/customers/slack", "type": "case-study", "date": "2026-07", "quote": "the team uses Claude Code to fix bugs and power teams to move faster." } ] } ] }, { "slug": "slingshot-ai", "name": "Slingshot AI", "website": "https://slingshotai.com", "what": "AI lab building foundation models for psychology.", "profile": "Slingshot AI runs multi-stage supervised fine-tuning and preference optimization pipelines using Together AI.", "facts": [ { "tool": "Together AI", "vendor": "Together AI", "layer": "training", "usage": "Runs supervised fine-tuning and preference optimization pipelines.", "status": "current", "confidence": "medium", "last_verified": "2026-07-27", "added": "2026-07-27", "evidence": [ { "url": "https://www.together.ai/customers/slingshot-ai", "type": "case-study", "date": "2026-07", "quote": "using Together AI for key components" } ] } ] }, { "slug": "sourcegraph", "name": "Sourcegraph", "website": "https://sourcegraph.com", "what": "Code intelligence platform with AI coding assistants.", "profile": "Sourcegraph partners with Fireworks AI for scalable model inference behind its real-time code assistance features.", "facts": [ { "tool": "Fireworks AI", "vendor": "Fireworks AI", "layer": "inference", "usage": "Provides scalable model inference for real-time code assistance.", "status": "current", "confidence": "high", "last_verified": "2026-07-27", "added": "2026-07-27", "evidence": [ { "url": "https://fireworks.ai/blog/story-sourcegraph-code-generation", "type": "case-study", "date": "2025-01-22", "quote": "“Fireworks has been a fantastic partner in building AI dev tools at Sourcegraph." } ] } ] }, { "slug": "spotify", "name": "Spotify", "website": "https://www.spotify.com", "what": "Music and audio streaming service.", "profile": "Spotify's fleet-management system, Honk, is the Claude Agent SDK running in Kubernetes pods against thousands of repositories, with CI verification on Linux and macOS in place of the LLM judge earlier versions relied on. Engineers work terminal-first in Claude Code across monorepos that run past 20 million lines, and the company attributes 73% of pull requests to AI authorship.", "facts": [ { "tool": "Claude Code", "vendor": "Anthropic", "layer": "ai-coding", "usage": "Terminal-first agentic coding across monorepos, with several background sessions.", "status": "current", "confidence": "high", "last_verified": "2026-08-22", "added": "2026-08-22", "evidence": [ { "url": "https://claude.com/customers/spotify-boris-office-hours", "type": "case-study", "date": "2026-08", "quote": "I run it in a bunch of Max sessions in a terminal" } ] }, { "tool": "Claude Agent SDK", "vendor": "Anthropic", "layer": "orchestration", "usage": "Powers Honk, the fleet-management agent automating code changes across repositories.", "status": "current", "confidence": "high", "last_verified": "2026-08-22", "added": "2026-08-22", "evidence": [ { "url": "https://claude.com/customers/spotify-boris-office-hours", "type": "case-study", "date": "2026-08", "quote": "It’s the Claude Agent SDK running in a Kubernetes pod." } ] } ] }, { "slug": "stably", "name": "Stably", "website": "https://stably.ai", "what": "AI-powered QA testing agents for web applications.", "profile": "Stably leverages Vercel's AI Gateway for AI scalability and large rate limits, and runs agentic code execution in Vercel Sandbox.", "facts": [ { "tool": "Vercel AI Gateway", "vendor": "Vercel", "layer": "inference", "usage": "Provides scalable model access with high throughput limits.", "status": "current", "confidence": "high", "last_verified": "2026-07-27", "added": "2026-07-27", "evidence": [ { "url": "https://vercel.com/customers/how-stably-ships-ai-testing-agents-in-hours-not-weeks", "type": "case-study", "date": "2026-02-17", "quote": "leveraging AI Gateway for AI scalability and large TPM limits" } ] }, { "tool": "Vercel Sandbox", "vendor": "Vercel", "layer": "guardrails", "usage": "Enables secure agentic code execution.", "status": "current", "confidence": "high", "last_verified": "2026-07-27", "added": "2026-07-27", "evidence": [ { "url": "https://vercel.com/customers/how-stably-ships-ai-testing-agents-in-hours-not-weeks", "type": "case-study", "date": "2026-02-17", "quote": "and Sandbox for agentic code execution." } ] } ] }, { "slug": "stack-ai", "name": "Stack AI", "website": "https://www.stack-ai.com", "what": "Enterprise platform for building AI agents without code.", "profile": "Stack AI chose Weaviate as the vector database powering its enterprise AI orchestration platform.", "facts": [ { "tool": "Weaviate", "vendor": "Weaviate", "layer": "rag-vector", "usage": "Vector database behind its enterprise agent platform.", "status": "current", "confidence": "high", "last_verified": "2026-07-27", "added": "2026-07-27", "evidence": [ { "url": "https://weaviate.io/case-studies/stack-ai", "type": "case-study", "date": "2026-07", "quote": "Stack AI chose Weaviate as the vector database to power their enterprise AI orchestration platform." } ] } ] }, { "slug": "stripe", "name": "Stripe", "website": "https://stripe.com", "what": "Payments infrastructure platform for the internet.", "profile": "A Stripe business value consultant prototyped a full customer-facing value calculator application with v0, which the team then shipped.", "facts": [ { "tool": "v0", "vendor": "Vercel", "layer": "ai-coding", "usage": "Prototypes customer-facing value calculator applications.", "status": "current", "confidence": "medium", "last_verified": "2026-07-27", "added": "2026-07-27", "evidence": [ { "url": "https://vercel.com/customers/how-stripe-built-a-game-changing-app-in-a-single-flight-with-v0", "type": "case-study", "date": "2026-01-28", "quote": "Mario Braz, Stripe GTM Business Value Consulting, used v0 to prototype a full application during a long-haul flight." } ] } ] }, { "slug": "substack", "name": "Substack", "website": "https://substack.com", "what": "Publishing platform for newsletters and podcasts.", "profile": "Substack moved both training and deployment of nearly all its ML models from AWS SageMaker to Modal.", "facts": [ { "tool": "Modal", "vendor": "Modal", "layer": "inference", "usage": "Runs ML training and deployment, replacing AWS SageMaker.", "status": "current", "confidence": "high", "last_verified": "2026-07-27", "added": "2026-07-27", "evidence": [ { "url": "https://modal.com/blog/substack-case-study", "type": "case-study", "date": "2024-05-20", "quote": "For nearly all these models Substack has moved both training and deployment from AWS SageMaker to Modal." } ] } ] }, { "slug": "suno", "name": "Suno", "website": "https://suno.com", "what": "AI music generation platform.", "profile": "Suno dynamically scales inference and batch pre-processing to thousands of GPUs with Modal.", "facts": [ { "tool": "Modal", "vendor": "Modal", "layer": "inference", "usage": "Scales inference and batch pre-processing to thousands of GPUs.", "status": "current", "confidence": "high", "last_verified": "2026-07-27", "added": "2026-07-27", "evidence": [ { "url": "https://modal.com/blog/suno-case-study", "type": "case-study", "date": "2024-02-21", "quote": "Suno uses Modal to scale inference and batch pre-processing to thousands of GPUs." } ] } ] }, { "slug": "terminal-x", "name": "Terminal X", "website": "https://www.terminal-x.ai", "what": "AI agent platform for investment managers.", "profile": "Terminal X rebuilt its retrieval architecture from the ground up with Pinecone at the core, powering real-time, context-aware retrieval across complex financial data.", "facts": [ { "tool": "Pinecone", "vendor": "Pinecone", "layer": "rag-vector", "usage": "Powers real-time retrieval across complex financial data.", "status": "current", "confidence": "high", "last_verified": "2026-07-27", "added": "2026-07-27", "evidence": [ { "url": "https://www.pinecone.io/customers/terminal-x/", "type": "case-study", "date": "2026-07", "quote": "Terminal X rebuilt its retrieval architecture from the ground up with Pinecone at the core." } ] } ] }, { "slug": "the-washington-post", "name": "The Washington Post", "website": "https://www.washingtonpost.com", "what": "US national news publisher.", "profile": "The Washington Post powers its public-facing AI journalism experience on Together AI, deploying open models on dedicated endpoints with hybrid serverless capacity for full model control.", "facts": [ { "tool": "Together AI", "vendor": "Together AI", "layer": "inference", "usage": "Serves open models behind a public AI journalism platform.", "status": "current", "confidence": "high", "last_verified": "2026-07-27", "added": "2026-07-27", "evidence": [ { "url": "https://www.together.ai/customers/washington-post", "type": "case-study", "date": "2026-07", "quote": "They deployed open models on Together AI with dedicated endpoints and hybrid serverless capacity, maintaining full model control and predictable pricing." } ] } ] }, { "slug": "the-weather-company", "name": "The Weather Company", "website": "https://www.weathercompany.com", "what": "Weather data and forecasting company behind weather.com.", "profile": "The Weather Company generates front-end interfaces with v0 to speed up its design-to-development cycles.", "facts": [ { "tool": "v0", "vendor": "Vercel", "layer": "ai-coding", "usage": "Generates front-end interfaces to speed design-to-development cycles.", "status": "current", "confidence": "high", "last_verified": "2026-07-27", "added": "2026-07-27", "evidence": [ { "url": "https://vercel.com/customers/how-the-weather-company-serves-real-time-forecasts-to-350-million-daily-active-users-on-vercel", "type": "case-study", "date": "2026-06-09", "quote": "web serving stack and CMS on Vercel and v0" } ] } ] }, { "slug": "trellix", "name": "Trellix", "website": "https://www.trellix.com", "what": "Cybersecurity company focused on extended detection and response.", "profile": "Trellix's professional services team built Sidekick, an internal agentic platform for log parsing and automation, with LangGraph and the visual LangGraph Studio, using LangSmith for experimentation and performance monitoring.", "facts": [ { "tool": "LangGraph", "vendor": "LangChain", "layer": "orchestration", "usage": "Builds the internal Sidekick agent platform for log-parsing automation.", "status": "current", "confidence": "high", "last_verified": "2026-07-27", "added": "2026-07-27", "evidence": [ { "url": "https://blog.langchain.dev/customers-trellix/", "type": "case-study", "date": "2025-04-21", "quote": "Trellix Professional Services Team used LangSmith and LangGraph – including the visual LangGraph Studio – to develop Sidekick" } ] }, { "tool": "LangSmith", "vendor": "LangChain", "layer": "evals-observability", "usage": "Monitors agent performance and debugs workflows through experiments and traces.", "status": "current", "confidence": "high", "last_verified": "2026-07-27", "added": "2026-07-27", "evidence": [ { "url": "https://blog.langchain.dev/customers-trellix/", "type": "case-study", "date": "2025-04-21", "quote": "Trellix used LangSmith for experimentation and to action upon performance metrics." } ] } ] }, { "slug": "trilogy", "name": "Trilogy", "website": "https://www.trilogy.com", "what": "Enterprise software operator running a portfolio of business software products.", "profile": "Fireworks AI became Trilogy's primary inference layer for internal deployment testing and early production workloads in its agentic workflows.", "facts": [ { "tool": "Fireworks AI", "vendor": "Fireworks AI", "layer": "inference", "usage": "Primary inference layer for internal agentic workflow deployments.", "status": "current", "confidence": "high", "last_verified": "2026-07-27", "added": "2026-07-27", "evidence": [ { "url": "https://fireworks.ai/blog/Trilogy", "type": "case-study", "date": "2026-06", "quote": "Over time, Fireworks became the primary inference layer for internal deployment testing and early production workloads." } ] } ] }, { "slug": "unify", "name": "Unify", "website": "https://www.unifygtm.com", "what": "Go-to-market platform with outbound AI agents.", "profile": "Unify relies on OpenAI's agentic capabilities and tool-calling to execute millions of automated research jobs and detect buying signals in real time.", "facts": [ { "tool": "OpenAI models", "vendor": "OpenAI", "layer": "models", "usage": "Executes automated research jobs and detects buying signals.", "status": "current", "confidence": "high", "last_verified": "2026-07-27", "added": "2026-07-27", "evidence": [ { "url": "https://openai.com/index/frontier-builders/", "type": "case-study", "date": "2026-07", "quote": "Unify relies on OpenAI’s agentic capabilities and tool-calling to execute millions of automated research jobs" } ] } ] }, { "slug": "valthos", "name": "Valthos", "website": "https://www.valthos.com", "what": "AI biodefense company for biological risk assessment.", "profile": "Valthos uses OpenAI as the reasoning layer tying together the biological models in its risk assessment pipelines.", "facts": [ { "tool": "OpenAI models", "vendor": "OpenAI", "layer": "models", "usage": "Reasoning layer for biological risk assessment pipelines.", "status": "current", "confidence": "high", "last_verified": "2026-07-27", "added": "2026-07-27", "evidence": [ { "url": "https://openai.com/index/frontier-builders/", "type": "case-study", "date": "2026-07", "quote": "Valthos uses OpenAI as the reasoning layer that ties together biological models" } ] } ] }, { "slug": "vanguard", "name": "Vanguard", "website": "https://www.vanguard.com", "what": "One of the world's largest investment management companies.", "profile": "Vanguard chose Pinecone as the vector database powering hybrid retrieval for Agent Assist, its customer-support assistant.", "facts": [ { "tool": "Pinecone", "vendor": "Pinecone", "layer": "rag-vector", "usage": "Powers hybrid retrieval for the Agent Assist support assistant.", "status": "current", "confidence": "high", "last_verified": "2026-07-27", "added": "2026-07-27", "evidence": [ { "url": "https://www.pinecone.io/customers/vanguard/", "type": "case-study", "date": "2026-07", "quote": "Vanguard chose Pinecone as their vector database to power hybrid retrieval for Agent Assist" } ] } ] }, { "slug": "vanta", "name": "Vanta", "website": "https://www.vanta.com", "what": "Trust management and security compliance automation platform.", "profile": "Vanta uses OpenAI models to reason over and translate data across its security and compliance platform.", "facts": [ { "tool": "OpenAI models", "vendor": "OpenAI", "layer": "models", "usage": "Reasons over and translates security and compliance data.", "status": "current", "confidence": "high", "last_verified": "2026-07-27", "added": "2026-07-27", "evidence": [ { "url": "https://openai.com/index/frontier-builders/", "type": "case-study", "date": "2026-07", "quote": "Vanta uses OpenAI models to reason and translate data across its" } ] } ] }, { "slug": "vercel", "name": "Vercel", "website": "https://vercel.com", "what": "Frontend cloud platform behind Next.js, v0, and the AI SDK.", "profile": "Vercel runs its v0 composite code-generation models and its Auto Fix model on Fireworks AI, using speculative decoding to speed up token generation.", "facts": [ { "tool": "Fireworks AI", "vendor": "Fireworks AI", "layer": "inference", "usage": "Runs v0 composite and Auto Fix models with speculative decoding.", "status": "current", "confidence": "high", "last_verified": "2026-07-27", "added": "2026-07-27", "evidence": [ { "url": "https://fireworks.ai/blog/vercel", "type": "case-study", "date": "2025-11-03", "quote": "Both Vercel’s Auto Fix model and its v0 composite model uses Fireworks’ Speculative Decoding to speed up token generation." } ] } ] }, { "slug": "vercept", "name": "Vercept", "website": "https://vercept.com", "what": "AI startup building vision-driven computer-use agents.", "profile": "Vercept deploys its custom computer vision models on Together AI's infrastructure for efficient, load-balanced inference.", "facts": [ { "tool": "Together AI", "vendor": "Together AI", "layer": "inference", "usage": "Deploys custom computer vision models via load-balanced inference.", "status": "current", "confidence": "high", "last_verified": "2026-07-27", "added": "2026-07-27", "evidence": [ { "url": "https://www.together.ai/customers/vercept", "type": "case-study", "date": "2026-07", "quote": "Together provides auto-scaling that monitors response latency and queue depth, enabling instant scaling without pre-provisioning." } ] } ] }, { "slug": "vodafone", "name": "Vodafone", "website": "https://www.vodafone.com", "what": "Multinational telecommunications company.", "profile": "Vodafone builds AI assistants and RAG pipelines for its data operations on LangChain components, constructs modular agents as LangGraph subgraphs, and tracks the LLM application lifecycle with LangSmith.", "facts": [ { "tool": "LangChain", "vendor": "LangChain", "layer": "orchestration", "usage": "Builds AI assistants and RAG pipelines for data operations.", "status": "current", "confidence": "high", "last_verified": "2026-07-27", "added": "2026-07-27", "evidence": [ { "url": "https://blog.langchain.dev/customers-vodafone/", "type": "case-study", "date": "2025-03-23", "quote": "“We’ve been using LangChain’s components for over a year now,”" } ] }, { "tool": "LangGraph", "vendor": "LangChain", "layer": "orchestration", "usage": "Constructs modular agents as subgraphs in multi-agent workflows.", "status": "current", "confidence": "high", "last_verified": "2026-07-27", "added": "2026-07-27", "evidence": [ { "url": "https://blog.langchain.dev/customers-vodafone/", "type": "case-study", "date": "2025-03-23", "quote": "Vodafone used LangGraph to construct modular agents as subgraphs" } ] }, { "tool": "LangSmith", "vendor": "LangChain", "layer": "evals-observability", "usage": "Tracks LLM application lifecycle, debugging, and evaluation.", "status": "current", "confidence": "high", "last_verified": "2026-07-27", "added": "2026-07-27", "evidence": [ { "url": "https://blog.langchain.dev/customers-vodafone/", "type": "case-study", "date": "2025-03-23", "quote": "With LangChain, LangGraph and LangSmith, Vodafone has successfully delivered advanced AI-driven solutions" } ] } ] }, { "slug": "xy-ai-labs", "name": "XY.AI Labs", "website": "https://xy.ai", "what": "Agentic AI for healthcare administration and revenue-cycle workflows.", "profile": "XY.AI migrated from a self-hosted stack to the Together Fine-Tuning Platform to train customer-specific Qwen models for explanation-of-benefits parsing, and serves them as testable endpoints on Together's serverless inference.", "facts": [ { "tool": "Together Fine-Tuning Platform", "vendor": "Together AI", "layer": "training", "usage": "Trains customer-specific Qwen models for EOB parsing.", "status": "current", "confidence": "high", "last_verified": "2026-07-27", "added": "2026-07-27", "evidence": [ { "url": "https://www.together.ai/customers/xy-ai-labs", "type": "case-study", "date": "2026-07", "quote": "XY.AI migrated from its self‑hosted stack to the Together Fine‑Tuning Platform" } ] }, { "tool": "Together AI", "vendor": "Together AI", "layer": "inference", "usage": "Serves fine-tuned models as endpoints for structured extraction.", "status": "current", "confidence": "high", "last_verified": "2026-07-27", "added": "2026-07-27", "evidence": [ { "url": "https://www.together.ai/customers/xy-ai-labs", "type": "case-study", "date": "2026-07", "quote": "Together AI turned fine-tuning, evaluations, and deployment into a repeatable loop from experimentation to generating a testable endpoint" } ] } ] }, { "slug": "yutori", "name": "Yutori", "website": "https://yutori.com", "what": "AI company building autonomous web-browsing agents.", "profile": "Together AI's inference platform is the backbone for Yutori's Scouts and Delegate products and its public Navigator API.", "facts": [ { "tool": "Together AI", "vendor": "Together AI", "layer": "inference", "usage": "Serves browser-use agents behind Scouts, Delegate, and Navigator.", "status": "current", "confidence": "high", "last_verified": "2026-07-27", "added": "2026-07-27", "evidence": [ { "url": "https://www.together.ai/customers/yutori", "type": "case-study", "date": "2026-07", "quote": "Yutori runs on Together AI, the AI Native Cloud" } ] } ] }, { "slug": "zapier", "name": "Zapier", "website": "https://zapier.com", "what": "Workflow automation platform connecting thousands of apps.", "profile": "Zapier's team uses Braintrust to log user interactions, dig into logs, track customer feedback, and manage test sets for AI evaluation.", "facts": [ { "tool": "Braintrust", "vendor": "Braintrust", "layer": "evals-observability", "usage": "Logs interactions, tracks feedback, and manages evaluation test sets.", "status": "current", "confidence": "high", "last_verified": "2026-07-27", "added": "2026-07-27", "evidence": [ { "url": "https://www.braintrust.dev/customers/zapier", "type": "case-study", "date": "2026-07", "quote": "The Zapier team uses Braintrust to log user interactions, dig into their logs, track customer feedback" } ] } ] }, { "slug": "zoominfo", "name": "ZoomInfo", "website": "https://www.zoominfo.com", "what": "B2B go-to-market intelligence and data platform.", "profile": "ZoomInfo adopted Pinecone as the vector database powering real-time semantic search for personalized contact recommendations.", "facts": [ { "tool": "Pinecone", "vendor": "Pinecone", "layer": "rag-vector", "usage": "Powers real-time semantic search for personalized contact recommendations.", "status": "current", "confidence": "high", "last_verified": "2026-07-27", "added": "2026-07-27", "evidence": [ { "url": "https://www.pinecone.io/customers/zoominfo/", "type": "case-study", "date": "2026-07", "quote": "ZoomInfo adopted Pinecone as the vector database powering the platform." } ] } ] } ] }