← ...
project 12: the feature store redis ml
project 12: the “feature store” (redis + ml)
scenario: the data science team needs “user average spend” for their model, with <10ms latency. the mission: build a feature serving layer.
- tech: redis, python api (fastapi).
- challenge: bridging offline (analytical) data to online (operational) apps.
- dev to prod:
- calculate “avg spend” in pyspark (project 4).
- push the result to redis: key=
user:123, value=500.20. - prod requirement: build a tiny fastapi endpoint
/get_feature/{user_id}that hits redis.