Building AI applications, agent workflows, and the engineering systems around them.
涓枃浠嬬粛 路 Explore my collections
I work across Python and TypeScript, from realtime voice applications and computer vision experiments to full-stack delivery and Linux automation. I care about clear interfaces, reproducible results, and tools that remain useful as models and providers change.
| Project | What you'll find |
|---|---|
| AI Engineering Workspace Standard | A minimal, agent-agnostic standard for durable engineering knowledge and handoffs. |
| Review Loop Skill (private) | A portable Codex / Claude Code workflow for independent architecture review, scoped implementation handoffs, and evidence-based verification. |
| EchoTalk | An English speaking practice application built around realtime voice conversations, LiveKit, FastAPI, and React. |
| DevOpsToolkit | Ansible-based Linux and WSL setup, with explicit boundaries between system, root, and user configuration. |
| Vite React Template | A reusable React and TypeScript starting point with testing, linting, and CI configuration. |
- Keep project knowledge portable across AI coding tools.
- Make changes small enough to review and verify independently.
- Treat tests, evaluation evidence, and deployment checks as part of delivery.
- Turn useful experiments into reusable tools; document the limits along the way.
Working stack: Python 路 TypeScript 路 React 路 FastAPI 路 PostgreSQL 路 Docker 路 Ansible
My repositories include maintained projects, experiments, learning material, and customized forks. Project READMEs describe their current scope and status; forks retain their upstream attribution.