I build and evaluate software across GPU systems, developer tools, and medical imaging. My work connects research experiments with interfaces, infrastructure, and evidence that people can inspect and use.
| Project | Engineering focus | Start here |
|---|---|---|
| TraceForge | Coding-agent trajectory analysis, evidence graphs, and a Jac/Python CLI | Project & sample workflow |
| openTorch / CGinS | PyTorch operator profiling, CUDA generation, and correctness feedback | Architecture & source |
| Coding-Agent Research Library | Literature discovery, source provenance, and thematic research indexes | Browse the library |
| Medical Physics Residency Site | Structured content, program navigation, and a Next.js interface | Application source |
Measure carefully. Preserve the workload, baseline, and evaluation context behind a result.
Make failures inspectable. Build tools that explain what happened and retain the evidence.
Finish the workflow. Connect models and experiments to usable interfaces and reproducible execution.
My broader research includes distributed model serving, coding-agent evaluation, and medical-image segmentation and source localization. These public repositories cover selected parts of that work.
Interested in research engineering, developer infrastructure, and systems software.

