I build real-time visual systems from physical and perceptual models.
Computer Vision PyTorch OpenGL / GLSL WebGL TypeScript
My current focus is the boundary between computation and visual experience: turning depth, motion, light, and atmosphere into interactive systems that can be observed, measured, and refined.
Real-time painterly rendering driven by monocular depth and optical flow.
- OpenCV capture and motion estimation → PyTorch/MPS depth inference → OpenGL/GLSL rendering
- A distilled depth model reduced parameters by 25.6x and improved application rendering from 52.9 to 134.6 fps on the measured setup
- Includes paired input/output recording, temporal-stability measurement, synthetic fixtures, and documented failed hypotheses
A real-time WebGL ray tracer for atmospheric scattering, volumetric clouds, and Fresnel ocean reflection. The sky is computed by numerically integrating light along each view ray.
- reslot — a production-minded appointment rescheduling reference app with transactional workflows, outbox delivery, tests, and public operations controls
- Auditloom — an evidence-based repository quality auditing CLI, designed around untrusted input, reproducible findings, and explicit verification boundaries
- Real-time computer vision and graphics
- Interactive systems grounded in measurable behavior
- Translating physical or perceptual ideas into working prototypes
- Architecture, testing, and documentation that make experiments reproducible