Built a real-time playing card detection system using a fine-tuned YOLOv26 model. Detects card rank and suit live via webcam with bounding boxes and confidence scores. Also includes a poker hand evaluator.

Researching self-supervised learning on binary classification of skin lesions (malignant vs benign) from dermoscopy images using DINOv2 and ResNet models on the MRA-MIDAS Stanford dataset.

Built a computer vision system in Python that processes parking lot video footage and classifies each space as occupied or empty on every frame using MOG2 background subtraction and Laplacian variance.

Built and shipped a temperature forecasting product from data preparation to cloud deployment. Owned model development, backend integration, and user-facing prediction workflows to deliver reliable forecasts in a simple web experience.