The infrastructure stack of AI.

The standard framework for building LLM-powered applications. LangChain's composable chain architecture became the de facto way developers orchestrate model calls, retrieval pipelines, and agent loops. Infrastructure that developers deploy in production and depend on daily.

Real-time AI assistance for customer-facing teams. Cresta's models process live conversations and surface guidance, knowledge, and scoring — infrastructure for contact center operations that compounds in value as the model learns from each interaction.

Purpose-built vector database for embedding storage and semantic search. As LLM applications proliferated, Pinecone became the default data layer — every retrieval-augmented generation (RAG) pipeline depends on a vector store, and Pinecone built the best one first.

ML experiment tracking, model registry, and observability platform used by ML engineers across thousands of teams. W&B is infrastructure: once integrated into a training pipeline, it becomes the system of record for every model iteration.

The central hub for open-weight models — hosting, sharing, and running inference. Hugging Face occupies the distribution layer of the AI stack: if you're not building a proprietary frontier model, you're likely discovering and deploying through them.

Cloud platform for running ML models via API. Replicate abstracts the complexity of model deployment — GPU provisioning, containerization, version management — into a clean developer interface. The picks-and-shovels layer for teams that build with models but don't want to run infrastructure.

Distributed inference compute for open-source LLMs. Together AI provides fast, cost-effective inference by aggregating GPU capacity across a decentralized network — infrastructure for teams that need frontier-speed inference without frontier-cost pricing.

AI-native search that combines LLM reasoning with real-time retrieval — the first production-quality alternative to the keyword search paradigm. Backed early when the search-with-LLMs thesis was nascent.