Your question is Latency vs Model Complexity. 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 do you balance the need for model complexity with the latency requirements of a production environment?
Explain how you would choose model architectures, allocate an inference latency budget, and decide what belongs in online versus offline computation. Discuss evaluation, cost, deployment, and monitoring considerations.