AI product prototypes · Installable agent skills · Inspectable decision tools
I turn messy asks into workflows you can run, skills you can install, and analyses you can check.
把业务问题收成可跑的流程、可安装的 Agent Skill、可复查的分析。
| Project | What you get |
|---|---|
| ai-product-portfolio | Case site for three AI product workflows — screenshots, demo video, Chinese PDF |
| text-rea | Installable writing skill: evidence → mechanism → transferable judgment |
| ai-video-preproduction-agent | Brief → concepts → script / storyboard → human-approved delivery package |
| safedrive-sql-analytics | MySQL model + business queries + research ledger you can re-run |
| choice-confidence-workbench | Offline confidence thresholds → release / review / labeled errors |
| table-questions-evidence | Excel-native Q&A with cell evidence and change checks |
- Name the decision and who owns it.
- Make the workflow explicit — inputs, human gates, failure modes.
- Ship something inspectable: prototype, skill, or analysis with a path to re-run.
Stack across projects: Python, SQL/MySQL, FastAPI, Next.js, agent skills (SKILL.md).
quant-alpha-research-framework · cn-equity-fund-evaluation · oregon-hie-ed-replication · index-futures-intraday-pairs-research · credit-default-cost-sensitive · campus-food-pickup-simulation · courier-action-classification