Your question is Latency vs Accuracy Trade-Offs. 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 do you approach the trade-offs between model latency and predictive accuracy in a production system?
Discuss how you would frame the latency and quality requirements, choose among model architectures or serving strategies, and decide what belongs online versus offline. Cover benchmarking, deployment, monitoring, rollback, and how you would validate that a latency optimization does not create unacceptable predictive degradation.