Projects & Applications
only some projects are posted; same w/github since i build fast and document slow.
for now, a slice of my best work.
Custom saber with a 3D-printed hilt, motion-reactive LED blade, and synchronized sound effects. The firmware reads the MPU6050 to trigger swing and clash patterns and uses a simple state machine for modes. Power is a protected Li-ion pack feeding a buck converter and MOSFET stage for stable LED current.
.jpg)
AI-powered anemia detection system using ML and computer vision to analyze fingernail color. Provides fast, non-invasive health assessment - users can monitor anemia risk anywhere with just a camera.
Winner of the Regional Award and Governor General's Award from Ingenious+ 2025.
Competition robot tuned for smooth line following and fast recovery using PID. I designed quick-swap attachments for pushing and tug challenges, and logged sensor data to iterate on gains. The focus was consistent lap times and predictable behavior across different track surfaces.
Low-cost 3-DOF arm with camera-based detection and basic inverse kinematics. I added a fast calibration routine and soft-limit safety to avoid joint overtravel. A Flask dashboard exposes jog controls, teach points, and simple status readouts.
MPH Data Learn
Hemorrhage protocol machine learning system using patient data to determine when to administer blood transfusions based on various clinical factors. Built during my studentship at UofT T-CarieM Research. Trained a Random Forest Classifier using Scikit-learn and an XGBoost Classifier using the XGBoost library, with hyperparameter tuning for optimal performance.
Cross-platform app with a clean flow from start to track to review, real-time stats, and minimal taps. I kept state management simple and responsive so the UI stays smooth on lower-end devices. Data is structured for weekly summaries and streaks.
Photo-first meal logging with ingredient recognition and macro estimates to cut logging time. The flow guides users to confirm or adjust results quickly and saves favorites for repeat meals. Emphasis is on fast capture, clear feedback, and simple history browsing.
A forearm armband that reads EMG or ECG signals and maps them to device or robot commands. I implemented basic filtering and envelope and threshold logic, with a desktop dashboard for live visualization and tuning. Bluetooth sends clean control events to downstream devices.