I’m a programmer based in San Francisco, USA, working across desktop AI, on-device intelligence, resilient mobile systems, and spatial computing. I like projects where the interface is only the visible edge of a deeper technical system—and where ambitious ideas are backed by measurable evidence.
- Grounded AI: Structured outputs, verification gates, local inference, and interfaces that make model behavior understandable.
- Resilient computing: Offline-first mobile systems designed for unreliable networks and real-world constraints.
- Spatial experiences: Reconstruction pipelines that turn images and evidence into explorable 3D environments.
- Native performance: Rust, C/C++, ARM-aware optimization, and careful boundaries between product code and systems code.
Explore all public repositories →
| Product layer | Systems layer | Intelligence layer |
|---|---|---|
| React, Next.js, Electron, Flutter | Kotlin, Rust, C/C++, BLE, SQLite | PyTorch, llama.cpp, ExecuTorch, multimodal APIs |
| Interfaces, accessibility, motion | Native workers, offline data, performance | Verification, reconstruction, local inference |
Language cards describe repository composition—not skill level. Public GitHub data is refreshed by the included workflow.
Pac‑Man replays the last year of contributions. The board regenerates every day.
Alternate arcade mode — Breakout
01 / Find the real constraint before polishing the apparent problem.
02 / Keep model output behind explicit contracts and verification.
03 / Measure performance on the hardware people actually use.
04 / Design failure states as carefully as the ideal path.
05 / Leave behind evidence: tests, benchmarks, provenance, and clear docs.
