Your question is Feature Store for Real-Time Inference. Take a moment with it on the right.
Talk me through your thinking if you like. When you're confident, submit your answer and I'll grade it like a real screen (7/10 or better passes).
How would you architect a feature store to support real-time inference for thousands of concurrent users?
Discuss the online and offline feature paths, storage and serving choices, freshness guarantees, consistency, and integration with model serving. Explain how you would handle feature computation, backfills, training-serving skew, feature drift, failures, monitoring, and capacity planning.
