TL;DR — ML Feature Store

Design an ML Feature Store that serves as the central repository for feature definitions, computation, and serving across an organization with 50+ ML models in production. The store must bridge the gap between offline training (batch, historical) and online serving (real-time, low-latency). Key features: Feature registry: define, version, discover, and document features. Offline store: point-in-time correct feature retrieval for training.