Projects

Built a scalar reverse-mode automatic differentiation engine from scratch in C++, implementing computation graphs, reverse topological traversal, gradient accumulation, arithmetic operations, power, ReLU, and tanh. Added automated testing with CTest and released v1.0.0.

GitHub ↗

Built a machine-learning model to forecast my bodyweight using a personal time-series dataset collected over ten months. Achieved an MAE of 1.14 lbs, outperforming a naive baseline of 1.49 lbs.

GitHub ↗

Developed a 3D deep-learning system for pulmonary nodule malignancy classification in CT imaging. Designed and trained an SE-ResNet3D model on LIDC-IDRI, outperforming a baseline 3D CNN on 5 of 6 evaluation metrics and improving ROC-AUC from 0.742 to 0.865.

Thesis ↗

Built an interactive physical model of the Burj Khalifa using computer vision and hand-gesture recognition to control lighting animations. The project was selected to represent RIT Dubai at Imagine RIT 2024 in Rochester, New York.