Some of my cool projects. You can see me yap about my projects on X!
Engineered a Python data pipeline to analyze 13 months of daily waste records (N=396). Identified 65% waste reduction (28.14 → 9.80 kg/day) and proved zero weekday variance via non-parametric testing. First author, COMSERVA Journal.
Assistive facial expression recognition mapping 478 MediaPipe landmarks to real-time intent classification (YES, NO, HELP, PAIN). Evaluated across Random Forest, Gradient Boosting, and Transformer models.
Browser-based macOS Sequoia-inspired desktop with custom window management, virtual web apps (Terminal, Safari, Spotify, Finder, VS Code), and Cloudflare Worker AI assistant.