jcorpac ML Feature Store Basics
The "Feature Store" is the bridge between data engineering and machine learning.
The Problem
"Training-Serving Skew": When the logic used to create features for training differs from the logic used in production.
Core Pillars
- Offline Store: For batch training on historical data.
- Online Store: Low-latency storage for real-time inference.
- Registry: A central catalog of feature definitions.
Tools
- Feast: The most popular open-source feature store.
- Tecton: A managed, enterprise-grade feature platform.
Best Practices
- Point-in-Time Correctness: Prevent leakage by only joining features as they existed at a specific timestamp.
- Reusability: Shared features across multiple models.