Describe the process of scaling model inference for a real-time API handling thousands of requests per second at RBC.