Feast

Feast is an open-source feature store designed for machine learning, providing structured data to AI and large language model applications at scale. It helps teams manage production ML systems by defining, validating, and serving features throughout the AI/ML workflow. Launched in 2018 through a collaboration between Gojek and Google Cloud, Feast addresses the challenges of feature engineering duplication across multiple ML teams.

The platform consists of two main components: an Offline Store for historical feature extraction and an Online Store for low-latency feature serving in production. Feast acts as a central hub for feature data and metadata, supporting various functions such as transformation, storage, serving, monitoring, and feature registry. It is particularly beneficial for ML platform and MLOps teams, including data engineers, software engineers, machine learning engineers, and data scientists, by enhancing collaboration and streamlining the productionization of real-time models.