A streaming feature pipeline computes rolling session features from an event stream and needs to keep a Unity Catalog feature table updated continuously. What does the Databricks documentation recommend for this workflow?
Answer & explanation
Correct answer: E
WHY E: Databricks documents a streaming feature computation pattern where a streaming DataFrame is passed to write_table for continuous updates. WHY NOT A: Serving endpoints are for inference, not direct feature ingestion. WHY NOT B: MLflow does not transform stream metrics into feature table writes. WHY NOT C: Streaming writes are supported. WHY NOT D: score_batch is for inference, not feature materialization.