Fourteen companies.
One stack.

An embeddable analytical query engine that runs complex SQL on local or remote data without a server process. Powers the next generation of data-intensive applications and notebooks.

High-performance vector similarity search engine designed for AI applications. Handles billion-scale embedding retrieval with built-in filtering and payload indexing.

Open-source framework for building, shipping, and scaling ML-powered services. Abstracts infrastructure complexity from model serving across cloud and on-premise environments.

Open-source AI-native database that stores objects and vectors simultaneously. Combines vector search with structured filtering for contextual AI retrieval at scale.

Python-native workflow orchestration platform for data and ML pipelines. Observable, recoverable, and developer-first — built to handle real production failure modes.

GraphQL data gateway that federates data sources at the edge. Enables AI applications to query multiple backends through a unified, performant schema layer.

Feature platform for real-time ML models. Lets teams define features in Python and serves them at millisecond latency — closing the gap between model development and production.

Indexing engine for unstructured data that updates incrementally rather than rebuilding from scratch. Critical infrastructure for AI systems that operate over live, evolving document corpora.

Cloud-native platform for running GPU-accelerated Python workloads without managing infrastructure. Handles cold-start latency, scaling, and dependency isolation — letting ML engineers ship models like functions.

Open-source vector database built on the Lance columnar format. Serverless, embeddable, and optimized for multimodal data — stores text, images, and structured metadata together with their embeddings.

Python-native feature engineering layer that unifies batch and streaming computation. Eliminates the training-serving skew problem by running the same feature definitions in development and production.

Open-source platform for building millisecond-latency data pipelines with exactly-once semantics. Bridges streaming and batch worlds through a unified collection model — operational CDC to analytical query in minutes.

Open-source data transformation framework built around Hamilton — a micro-framework for defining feature and data transformations as directed acyclic graphs. Automatic lineage, testability, and versioning built in.

Change-data-capture platform that streams transactional database updates to analytical warehouses in near real-time. Eliminates nightly batch loads and keeps analytics current within seconds of operational writes.