Understand how model inference works, then design a provider gateway with failover and backpressure for the Model Routing and Inference interview.