LLM Serving & System Design Doubt Session with Harshal Dharpure

Bring your questions on LLM serving and inference system design: paged attention, continuous batching, prefill vs decode, chunked prefill, quantization choices, speculative decoding or sizing a deployment for a latency and cost target. We work through them on a whiteboard with real numbers. You leave unstuck, with the reasoning written down so you can explain it in an interview.