Your question is Batch GPU Inference API. 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).
Design a high-throughput Batch GPU Inference API capable of handling 100,000 requests per second across heterogeneous cloud providers.