Whenever you're ready, walk me through how you'd approach this.
Describe how you would deploy a model to a production environment to ensure low latency.
Explain the serving architecture, model optimization techniques, online versus batch inference choices, and how you would measure latency and reliability. Address deployment validation, scaling, monitoring, rollback, and behavior when the model or its dependencies are unavailable.