Companies we've backed

Durable workflow orchestration for modern applications — step functions, background jobs, and event-driven pipelines that are observable and recoverable by default. Built for the AI-native application era.

A headless, framework-agnostic rich-text editor built on ProseMirror. Powers collaborative editing and AI-generated content surfaces in B2B SaaS products. Developer primitive, not an app.

Cube provides a universal semantic layer that sits between data warehouses and downstream applications. Lets engineering teams define metrics once and expose them consistently across BI tools, embedded analytics, and AI applications.

Guardrails AI provides an open-source framework for validating LLM outputs against structural and semantic constraints. Enables engineering teams to wrap LLM calls with type-safe validators, hallucination checks, and policy enforcement.

Change-data-capture and streaming infrastructure for Postgres. Captures every database mutation and routes it to downstream consumers — queues, search indexes, AI pipelines — without polling or custom triggers.

A platform for authoring, sharing, and running codemods at scale. Automates framework upgrades, API migrations, and large-scale refactors across enterprise codebases — saving engineering weeks per migration cycle.

AI-native workflow automation purpose-built for finance operations teams — AP/AR, close processes, variance analysis. Replaces spreadsheet-and-email workflows with durable, auditable, AI-assisted execution pipelines.

An AI gateway and observability platform sitting between application code and LLM providers. Routes requests, enforces policies, logs every call with semantic context, and provides cost and quality analytics across models and providers.

Humanloop provides prompt management, evaluation, and monitoring infrastructure for teams building LLM-powered features. Combines version control for prompts with systematic evaluation workflows and production monitoring.

Digital contract lifecycle management platform that uses AI to automate contract creation, negotiation, and analysis. Enables legal and procurement teams to move from static documents to structured, auditable workflow with built-in compliance guardrails.

Tines automates repetitive manual processes in security operations, IT, and engineering through a no-code workflow builder. Connects to enterprise tools via APIs and enables teams to build automation without engineering resources.

Merge provides a single API to access data from hundreds of HR, ATS, CRM, accounting, and ticketing platforms. Eliminates the need for B2B SaaS companies to build and maintain individual integrations with each customer's tech stack.

A hybrid vector-relational store purpose-built for enterprise RAG applications. Combines dense vector search with metadata filtering and access-control scoping, so retrieval results respect the same permission model as the underlying data.

Baseten provides infrastructure for deploying and scaling ML models in production. GPU-optimized serving with auto-scaling, model versioning, and monitoring — designed for teams that need production inference without managing Kubernetes.

Workflow infrastructure for AI pipelines that require human review, approval, or intervention at defined checkpoints. Provides a structured handoff layer between automated AI steps and the people accountable for the final decision — without breaking the workflow.

Open-source infrastructure for ingesting and preprocessing unstructured data — PDFs, emails, images, HTML — into clean, chunked formats ready for LLM pipelines and RAG applications. The ETL layer between raw enterprise documents and vector stores.