Build versioned feature definitions, point-in-time historical retrieval and online serving with measurable parity.
Learning path
- Feature-store contract: entity keys, event time and availability
- Historical feature retrieval: point-in-time joins without future leakage
- Feature definitions: versioned views, ownership and lineage
- Online feature serving: materialization, staleness and fallback
- Feature parity: compare historical replay with online observations
- Feature operations: backfills, deletion and atomic publication
- Project: build a point-in-time receipt-triage feature service
Connected foundations
Use the earlier measurement and data contracts as prerequisites.
Further paths
Practice
Continue into another subject: Privacy-Aware ML Tutorial.