Design a serving architecture that supports both a 20ms fraud check and a nightly portfolio score.
PICTURE THIS: DATA SPLIT
Fit on train, tune on val, report on test once.
Simple meaning
I would register one model family but two serving paths: an online path with an in-memory or Redis feature store, HPA, and a strict timeout plus heuristic fallback.
WHY — Train vs Serve instead of guessing?
Why interviewers care about Train vs Serve:
separate people who only read docs from people who shipped.
and tied to MLOps work.
Name the idea, why it exists, then one short example.
End with when you use it and one common pitfall.
STEPS — What happens step by step?
Before you speak the answer, walk the interviewer through these steps:
- 1I would register one
model family but two serving paths: an online path with an in-memory or Redis feature store, HPA, and a strict timeout plus heuristic fallback.
- 2The batch path would
run as a scheduled job on cheap compute writing to the warehouse with the same transform library.
- 3Shared evaluation and a
registry alias keep the two paths from drifting, while capacity and SLOs stay independent.
- 4Give an example
One tiny concrete case you can say aloud.
- 5Common mistake
What juniors usually get wrong.
- 6Close
When you pick this over the alternative.
EXAMPLE — See it in action
Here's a short line you can speak, broken into clear beats:
Note: Adapt this scaffold to your own project — keep it under 60–90 seconds.
Key takeaway
I would register one model family but two serving paths: an online path with an in-memory or Redis feature store, HPA, and a strict timeout plus heuristic fallback. The batch path would run as a scheduled job on cheap compute writing to the warehouse with the same transform library.