Monitoring with delayed ground truth
Design a monitoring system for a loan-default prediction model where the true label — did the borrower actually default — only resolves months to years after the prediction. You can't wait that long to know the model is degrading. Walk through how you monitor model health when ground truth is delayed by quarters, what early proxy signals you use, and how you avoid both crying wolf and missing a real regression in the meantime.
What a strong answer looks like
Clarify scale and constraints first. Propose a clean component breakdown, then go deep on the hard parts (data model, bottlenecks, consistency, failure modes) and name the trade-offs you are making.
Clarify5:00 left
Estimate5:00 planned
Design15:00 planned
Deep dive12:00 planned
Failure8:00 planned
Which questions mattered is sealed until you submit. Telling you now would just be handing over the edge cases.
Run or narrate your approach, then ask the coach.