I build reliable AI agent systems and the developer tools that make them easier to understand.
In August 2026, my pull request openai/openai-agents-python#4363 was merged into OpenAI's Agents SDK for Python.
Fixed the AnyLLM Chat Completions adapter so
parallel_tool_callsis forwarded only when converted tools exist—including handoff-only agents—and added a six-case regression test matrix.
It was a focused reliability fix, but a deeply meaningful milestone for me: code I wrote now helps improve an open-source AI agent framework used by developers around the world.
| Project | What it does |
|---|---|
| traceweave | Distributed tracing and observability for AI agents. Follow LLM calls, tool use, handoffs, and agent decisions from a TUI or web dashboard. Available on PyPI. |
| agentmop | A local-first CLI that finds, explains, and safely cleans files left behind by AI coding agents—with instant restore. |
| smart-hand | An AI-powered robotic hand controlled by voice, with an LLM reasoning about finger positions in real time. |
- 🧠 Focused on AI agents, multi-agent systems, observability, and LLM reliability
- 🚀 Currently building products and engineering systems at a startup
- 🌱 Active open-source contributor across the AI developer ecosystem
- 💬 Previously a software engineer at WeChat, Tencent
- 🎓 B.Eng. in Software Engineering, Wuhan University (2019–2023)
- 📍 Based in Guangzhou, China


