You have a model in production, and the team wants a clear way to catch feature drift and model quality issues as they happen. You need a monitoring setup that can separate normal variation from meaningful degradation, and that can trigger alerts without overwhelming the oncall engineer.
Design a monitoring and alerting system to detect feature drift and model degradation in real time.