Your question is LLM Serving for High 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).
Explain the architectural considerations for system design for LLM serving at high latency requirements.
Discuss how you would design the online serving path, manage variable-length generation, scale compute, and balance latency, quality, and cost. Cover batching, caching, model placement, streaming, observability, and failure handling.