Computer engineering student. I build things that work on real hardware — multi-camera capture, gesture recognition, fall detection — not just notebook demos.
FERN — real-time foot gesture recognition
MediaPipe skeleton → CNN-BiLSTM-Attention → 7-class output
93.67% test accuracy across 35 subjects, 100% skeleton detection rate, 823K params, ~20ms/window on an RTX 3070. Live roadmap tracked at Vision-Orchestration.github.io/FERN.
DroidGrid — multi-phone camera controller
Preview, record, and snapshot from up to 10 Android phones simultaneously over DroidCam. Self-healing streams, non-blocking I/O, inline session editor.
Folio Finder AI — fall detection from video
R(2+1)D-18 trained on a custom dataset of ~7,000 clips. 98.71% F1 on a held-out test set. Full training pipeline with AMP, class weighting, and early stopping.
TODO — terminal task manager
Ultra-light CLI task manager. Nested tags, priorities, due dates, optional encryption, per-user data. Lives entirely in your terminal.
Vision / ML PyTorch MediaPipe OpenCV R(2+1)D CNN-BiLSTM-Attention
Languages Python C/C++ (Arduino) Bash
Systems CUDA ffmpeg SQLite Git
Hardware RTX 3070 · Ryzen 7 5800H · 32GB RAM
