About
I am a software engineer at Huawei and a founding contributor to vLLM-Omni, the open-source any-to-any multimodal LLM serving framework under vllm-project — I built the initial engine core and the AR engine, review multimodal engine PRs, and single-handedly perform its recurring upstream vLLM rebases. I also built vllm-omni-copilot, a LangGraph multi-agent system that automates the rebase end-to-end.
Before Huawei I studied computer science at UCLA (M.S.) and HKUST (B.Eng.). Outside work I pursue independent research on object-centric generative world models.
Skills
PythonCI/CDMulti-agent orchestration (LangGraph)