I build software at the boundary of machine learning reliability, quantitative decision-making, and native Apple/Linux tools. I study Mathematical Sciences and Financial Engineering at Washington University in St. Louis.
- ItoCanvas — offline macOS options laboratory for Black–Scholes–Merton pricing, Greeks, implied volatility, strategies, and scenario analysis.
- DualTyper — privacy-first macOS menu-bar translation for explicitly selected text, using Apple’s on-device Translation framework.
- numguard — high-precision checks for numerical ML kernels, with independent reference calculations and adversarial fixtures.
- qdrift — exact-oracle checks for affine INT8 quantization arithmetic and round-boundary drift.
- recallwatch — detects tail-query recall collapse in approximate-nearest-neighbor indexes.
- privaudit — local Linux microphone and camera access history built on PipeWire.
- Numerical correctness, silent failure detection, and reproducible ML experiments.
- Tools that work offline, explain their limits, and fail safely.
- macOS and Linux behavior verified on the operating system where it matters.
- Small, focused software with real tests rather than claims of completeness.
Each project README contains installation instructions, examples, verification status, and known limitations. I’m open to conversations about software engineering, quantitative finance, ML infrastructure, and analytical product work.