Software Engineer · Huawei2025 – present
Founding contributor to vLLM-Omni, the open-source multimodal (any-to-any) LLM serving framework under vllm-project: built the initial engine core — executor, worker, model runner — in the repo's first week
Implemented the AR (autoregressive) engine: AR GPU model runner & worker, scheduling components, Omni model abstractions, and multimodal output processing
Designed engine-level KV-cache management for the Block Diffusion Engine / DreamZero world model (RFC #4534) and authored the VGGT engine RFC, extending vLLM-Omni to world-model serving
Single-handedly performed every recurring upstream vLLM rebase (through v0.24.0): dependency and API alignment across 8 engine modules, wheel/CI pinning, and post-rebase review follow-ups (PR #4830)
Built vllm-omni-copilot, a LangGraph multi-agent system that automates the rebase end-to-end — parallel per-module LLM agents with plan-review gating, local pipeline testing, autonomous Buildkite CI debugging, and a self-improving skills/debug-memory knowledge layer
Core reviewer for multimodal engine PRs: audio talkers (CosyVoice3, MOSS-TTS, Higgs-Audio), image generation (HunyuanImage3), XPU enablement; shipped production fixes for concurrent PCM audio streaming (issue #4411)