I build quantitative research infrastructure and evidence-led software at Canli Capital.
My current focus is ALPHAC: an open multi-asset research and portfolio platform designed around point-in-time data, realistic execution, explicit trial accounting, reproducible experiments, and publication of negative results. The objective is not to advertise a backtest; it is to make every published claim inspectable and every limitation difficult to hide.
I am also building TraceAxiom, a verification-first AI software engineering system that coordinates specialist coding agents and binds accepted changes to independent build, behavior, accessibility, security, and authority evidence.
| Project | Purpose |
|---|---|
| TraceAxiom | Verification-first AI software engineering, dependency-aware specialist agents, independent evaluation, and source-bound evidence. |
| ALPHAC | Python research engine, validation framework, execution simulation, portfolio construction, and machine-readable evidence artifacts. |
| Canli Capital | Open research and evidence site for ALPHAC's methodology, paper record, kill log, corrections, and technical writing. |
| CubeDuel | Speedcubing platform with server-verified solves and a from-scratch sub-second Kociemba solver. |
| Hollowtide | Deterministic browser-game simulation with a measurement harness and zero external asset files. |
- Walk-forward and purged validation before admission claims
- Point-in-time lineage and untouched holdouts
- Deflated Sharpe, PBO, sensitivity, capacity, and regime tests
- Realistic costs, liquidity constraints, market status, and operational failure modes
- Trial ledgers, kill criteria, corrections, and negative-result publication
ALPHAC and Canli Capital currently publish research and paper-trading evidence. Nothing in these repositories is investment advice, an offer, or a representation of guaranteed performance.