Feature store

A feature store is a centralized repository used in machine learning to store, manage, and serve features for model training and inference.¹ It provides a unified interface for data scientists and engineers to access curated, reusable features derived from raw data, ensuring consistency between training and production environments.² Feature stores typically support batch and real-time data pipelines, enabling efficient feature computation, storage, and retrieval at scale.

Feature stores play a critical role in operationalizing machine learning systems by improving reproducibility, reducing data leakage, and promoting collaboration across teams.³ They often have features like feature versioning, metadata management, and access control that help keep data quality and governance high.⁴