Your question is Real-Time Inference Latency. 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).
How would you optimize inference latency for a real-time trading model at Citi?
Explain how you would profile the end-to-end path, choose model and hardware optimizations, and preserve prediction quality and trading safety. Cover online versus batch work, capacity planning, monitoring, fallbacks, and the distinction between model latency and downstream order execution latency.