AI Application & Full-Stack Engineer · Open-Source Contributor
I build digital-human workflows, LLM/agent systems, and developer tools with a focus on reliability and maintainability.
专注数字人、LLM / Agent 工作流与开发者工具，坚持用测试和工程质量把想法做成产品。
- AI applications: self-hosted digital humans, RAG, LLM agents, and human-in-the-loop workflows.
- Developer experience: diagnostic tools, automation, reproducible builds, and evidence-first debugging.
- Open source: focused, test-backed contributions to active AI and developer-tooling communities.
| Project | What it demonstrates | Core stack |
|---|---|---|
| HeyGemWeb | A production-oriented Gradio UI and workflow layer for self-hosted HeyGem-compatible digital-human services. | Python · Gradio · FFmpeg |
| Command Frontline | An original natural-language RTS with deterministic simulation, command planning, tests, and accessible controls. | TypeScript · React · Canvas · Playwright |
| Repo Doctor | An evidence-first CLI for diagnosing repositories that will not build or start. | Java 21 · Picocli · JUnit |
| Slop Zoo Cannon Game | A playable 3D arcade vertical slice with Blender assets, performance budgets, safe desktop packaging, and resilient saves. | Three.js · Blender · Electron |
| AI Playground | Ten progressive projects covering RAG, tool use, LangGraph, persistence, HITL, and multi-agent orchestration. | Python · LangChain · LangGraph |
Recent merged work includes contributions to Tencent/WeKnora, OpsiMate, and canvas-api-mcp. I also contribute to larger AI engineering projects such as MLflow and OpenHands.
I prefer small, reviewable changes backed by reproductions and regression tests, then grow toward deeper component ownership.
日拱一卒 · Ship useful software · Contribute upstream