Your question is Observability for Production LLMs. 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 approach monitoring and observability for a production LLM service to detect drift or degradation?
Explain the signals, dashboards, alerts, evaluation loops, and rollback mechanisms you would use. Cover infrastructure health, model behavior, data and feature drift, safety, cost, latency, and the challenge of evaluating outputs when labels are delayed or unavailable.