Your question is Managing KV Cache for LLMs. 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 manage KV-cache in large language model inference systems?