Your question is Inference Speed vs Accuracy. 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 a model for inference speed without sacrificing significant accuracy?
Describe and implement a practical optimization workflow for a supervised classification model. Your solution should profile the current implementation, compare faster model candidates, evaluate accuracy and latency together, and justify the final accuracy-latency tradeoff. Include production considerations such as batching, feature preprocessing, memory usage, hardware, and monitoring.