Computer Science at Durham · Agent infrastructure · Quant systems
I build tools that route work between AI and people, test trading ideas, and improve the open-source software I use.
| Project |
|---|
 warrant · live demoRoutes work between humans and AI using evidence-based scores. Irreversible actions need signed human approval, and outcomes improve future routing. Built solo; top 5 of 53 at GenAI Fund Agentic AI Build Week 2026. |
 cortexChooses the workflow, specialist, and effort tier for coding tasks, then logs decisions and learns from outcomes. |
| 📈 multi-factor-equity-screener A public write-up on forward-testing a trading screener. Pre-registered checks caught an inverted ranking model; the implementation is private. |
- Data engines: DataFusion (Spark decimal semantics), Arrow Rust (cast-module refactor), and Ballista (benchmark failure artifacts).
- Agent tooling: FastMCP (resource subscriptions) and mcp-atlassian (Jira parent updates, Jira formatting).
- Developer tools: Vercel Chat (Slack attachments), Scrapling (cached browser cookies), Understand-Anything (hook input), last30days, git-proxy, and toolport.
Languages
Frameworks, libraries and tools
Before Durham, I was a 3rd Sergeant in the Singapore Army's Military Police and led a 10-person team. Outside the terminal, I run long distances, take street photos, play guitar, and care a little too much about coffee.