Your question is Inference Under Constraints. 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 manage memory and compute constraints when running inference on large models?
Discuss how you would profile the workload, select optimization techniques, design online and batch inference paths, and preserve quality under resource limits. Cover deployment tradeoffs, capacity planning, monitoring, and failure recovery.