Infrastructure · Established · Advanced
Feature Store
A shared system for computing, storing and serving model input features consistently between training and inference.
What Feature Store is
The classic production bug is training-serving skew: a feature computed one way offline and another way online. A feature store removes that by defining each feature once.
How it works
Feature definitions produce both a historical table for training with point-in-time correctness and a low-latency online store for serving, with lineage and monitoring attached.
Why it matters
It makes features reusable across teams and eliminates a whole category of silent accuracy loss.
Common uses
- →Real-time fraud and risk scoring
- →Recommendation features
- →Shared enterprise ML platforms
Strengths
- ✓Consistency across environments
- ✓Feature reuse
Watch for
- ✓Heavy for small teams
- ✓Adds a platform dependency
Continue exploring
More in this collection
Browse all AI Concepts