Building research-grade systems where markets, data, and AI meet.
Reproducible by default · point-in-time honest · auditable end to end
| Domain | Tooling |
|---|---|
| Languages |      |
| Data & Quant |        |
| Engineering |        |
| AI & Agents |     |
| Project | What it does | Stack | ★ |
|---|---|---|---|
| axiom-foundry | Research-grade quantitative investing workbench — official-source macro data, point-in-time-aware factor research, backtesting, and auditable reports. | Python FastAPI MCP Qlib |
|
| AnalystCollective | Story → Numbers → Value. A Damodaran-method equity valuation engine: DCF, Monte Carlo, reverse DCF, SOTP, comps, and archetype playbooks. | Python |
3 |
| deep_option | HK/US equity options research and risk agent — a source-available prototype wired to live brokerage and market data. | Python futu-api OpenBB |
2 |
| zhouyi-ai | AI-assisted Yijing toolkit — hexagram casting, Four Pillars charting, five-element analysis, and reading generation. Live demo | TypeScript |
6 |
| dsh-plugin-updater | Plugin update center for DeepSeek Harness — version tracking, integrity verification, atomic replacement, hot reload. | TypeScript Cordis |
|
| da-xiao.skill | A reusable AI skill that gives an agent the full persona of a veteran A-share market commentator — voice, judgment, and temperament, not just quotes. | Claude Code Skill |
1 |
| dongfangbai | Personal digital garden — a student writing about markets and history. | Astro Cloudflare |
- Point-in-time discipline — survivorship-bias-free factor pipelines where every input is timestamped to what was actually knowable
- Agentic research workflows — MCP servers, reusable skills, and tool use that stays auditable rather than magical
- Valuation as a craft — DCF, reverse DCF, scenario trees, and Monte Carlo as ways of asking better questions
- Writing — markets and history, at the pace of reading rather than the pace of news
Data over narrative. Methods over opinions.