https://tarunajjarapu.github.io/writing/ Tarun Ajjarapu — ML systems engineer working on infrastructure and agents at Judgment Labs. Technical writing and videos on systems software and ML infrastructure. en-us https://tarunajjarapu.github.io/writing/inside-an-inference-runtime/ https://tarunajjarapu.github.io/writing/inside-an-inference-runtime/ How vLLM turns requests into GPU work through continuous batching, KV-cache paging, and token scheduling. ML Sun, 01 Nov 2026 00:00:00 GMT