Feast
lastmod 2025-12-30created 2025-12-303 edits
Feast (Feature Store) is an open-source feature store that helps teams operate production ML systems at scale by allowing them to define, manage, validate, and serve features for production AI/ML.
Components
Feast's feature store consists of two foundational components.
- An offline store for extracting historical features used for model training.
- An online store for serving features with low latency to production systems and applications.
Feast is a configurable operational data system that reuses existing infrastructure to manage machine learning features and serve them to real-time models.
Features
Specifically, Feast provides the following.
- A Python SDK for programmatically defining features, entities, sources, and (optionally) transformations.
- A Python SDK for reading and writing features from configured offline and online data stores.
- An optional feature server for reading and writing features (useful for languages other than Python)
- A user interface (UI) for viewing and exploring information about the features defined in a project.
- A CLI tool that can be used to view and update feature information.
What you can do
With Feast, ML platform teams can do the following.
- Make the features needed for training and low-latency serving consistently available. Manageable items are:
- An offline store (for processing historical data for batch scoring or model training)
- A low-latency online store (to support real-time prediction)
- And a proven feature server (serving precomputed features online)
- This prevents future feature values from leaking into models during training.
- This ensures model portability when moving from training models to serving models, from batch models to real-time models, and from one data infrastructure system to another.