MLOps Academy · Lesson

Materialize Features to an Online Store

Push features so they are ready at low latency.

Why Materialize

Your features sit in batch storage, but predictions need them in milliseconds. Materializing copies the latest values into a fast online store. ⚡

Offline Is Too Slow

The offline store is great for history but scanning Parquet at request time would be far too slow for live serving.

All lessons in this course

  1. Why Feature Stores Exist
  2. Define Feature Views with Feast
  3. Materialize Features to an Online Store
  4. Fetch Point-in-Time Correct Features
← Back to MLOps Academy