Waypoint · Research ▲ 1,400 ft
Three projects, one thread. I like teaching computers to see things people have to squint at: a detail in a photo, a cavity, or one whale's tail fluke among fifteen thousand.
Accepted · IEEE BigData 2026
SDSU Data Science Lab, with Prof. Hajar Homayouni
Can a model answer questions about an image better if vision and language attend to each other equally, or if one leads? I trained and compared both on VQA v2, then figured out why the asymmetric version overfit.
Read the full write-up →
Provisional patent filed · Aug 2026
Independent invention, sole inventor
My dentist still checks teeth with a mirror and a light. I designed a toothbrush-sized scanner with multi-spectral lighting and a camera, plus a per-tooth AI model that looks for decay and gum disease.
Read the full write-up →
Published · NHSJS, Sep 2025
Independent research, mentored by Arpit Jasapara (UCLA)
Telling individual whales and dolphins apart from photos of their flukes and fins. EfficientNetB5 with an ArcFace loss reached a MAP@5 of 0.88 on the Happywhale dataset.
Read the full write-up →