AI tools, developer infrastructure, and research experiments. Interested in the intersection of language models and real-world systems.
I build tools that treat language models as primitives, not products — gateways, agents, skills, and pipelines that other developers can compose.
Most of my work lives at the seam where infrastructure meets cognition: routing requests across model providers, giving agents persistent memory, turning long-form research into reproducible loops.
Outside of code I read about materials science, music theory, and Kahneman-style decision making. I write irregularly on X about whatever I'm currently confused by.
AI API Gateway Platform
Subscription quota distribution for AI APIs. Multi-account management, load balancing, request forwarding — built with Go + Vue. Trending on GitHub.
Your Second Digital Mind
iOS note app with four cognitive layers — Inbox, Mind, Insight, Ground. Expo (React Native) + FastAPI backend deployed on Railway.
Claude Code × 16 Platforms
Claude Code skill bridging 16 social platforms via CLI — Bilibili, Zhihu, X, YouTube, Reddit and more. Reuses Chrome sessions, zero API keys needed.
Materials Science AI Agent
AI coding assistant purpose-built for VASP computational materials science at Nankai University. Customised from OpenCode with domain-specific tools and agents.
Autonomous AI Research Loop
Give an AI agent a training setup and let it run overnight. It modifies code, trains 5 min, evaluates, iterates — you wake up to a log of experiments.
Have a project, a question, or just want to say hi?