I build tools for local AI workflows and data analysis, with an emphasis on clear boundaries, inspectable outputs, and reproducible checks.
| Project | What it explores | Current scope |
|---|---|---|
| FitTwin · Live demo | A browser-local body profile and garment-sizing toolkit with TypeScript SDK and local MCP tools | Alpha; manual measurements, transparent sizing rules, optional AI companion |
| Delphi Energy | A Python CLI for structured, multi-role decision review | Experimental; model critique and an inspectable ledger, not independent fact verification |
| Commerce Analytics | SQL/Python analysis and a bilingual Streamlit dashboard for an apparel-commerce scenario | Synthetic data; explicit metric definitions and reproducible analysis |
关注本地 AI 工具、可解释的分析过程与可复现的工程实现。项目文档会区分已经实现的能力、验证结果和现有限制;电商分析使用合成数据,AI 评审结果不等同于事实核验。