BioReact-Pi
Edge-AI bioreactor controller on a Raspberry Pi that reads a live temperature sensor, feeds it into a logistic growth model to predict bacterial biomass, and streams everything to a live web dashboard — including an animated petri-dish colony visualization. Built at CU Hacking 2026 for the QNX Hackathon Challenge, winning 3rd Place.
Northern Shift Guard
An explainable AI safety system for Northern Ontario mining sites that detects PPE non-compliance and fatigue risk, checks compliance against zone-specific requirements, reasons over evidence with Nemotron, and stores every flagged decision in a local audit log — no black boxes. Won 2nd Place and Best Use of NVIDIA at Cursor Hackathon Sudbury 2026.
MalariAI
Deployed on Hugging Face Spaces via Flask. Upload a Giemsa-stained blood smear image and MalariAI automatically segments cells, identifies infected ones, calculates the infection rate, and determines the dominant parasite stage — all with Grad-CAM++ visualizations for explainability. Backed by a two-stage pipeline: watershed-based segmentation and EfficientNet-B0 classification trained on NIH BBBC041 and MP-IDB datasets.
