What it is: Feast is the leading open-source feature store for machine learning, with over 12M downloads. MongoDB is now a natively supported online and offline store in Feast. This enables ML teams to serve features at low latency and generate point-in-time-correct training datasets directly against MongoDB. The integration also includes MongoDB Vector Search support for similarity search over feature embeddings.

Who it's for: ML engineers, data scientists, and MLOps practitioners building production ML pipelines on Feast - particularly teams whose operational or application data already lives in MongoDB.

Why it matters: A typical Feast deployment runs three databases: the application's primary database, a dedicated online store for low-latency serving, and a separate warehouse for offline retrieval. By using MongoDB as both the online and offline store in Feast, teams can consolidate feature storage, real-time serving, and vector search onto a single Atlas cluster - eliminating ETL pipelines and the operational overhead of fragmented ML infrastructure. Feature data lives next to the operational data it was derived from, accelerating feature iteration from training to production inference.

How to get started: Check out the documentation to learn about integrating MongoDB with Feast.