My Projects
A collection of things I've built, learned from, and am proud of. From ML experiments to full-stack applications.
Built a PyTorch federated learning simulator to stress-test FedAvg with non-IID data, partial participation, and async client delays. Found MNIST remains stable while CIFAR-100 accuracy collapses when client drift and staleness combine.
Built an AR shopping assistant for the Mentra Hackathon that prevents overpaying. Uses computer vision and AI to instantly recognize products and compare prices across retailers through voice commands on smart glasses. Gained major social media traction and recognition from Mentra's leadership.
A decentralized drone delivery platform with a Go routing engine using goroutines, worker pools, and grid-based shortest paths. Dispatches fleets across cities without tying merchants to droneports, with a TypeScript/Convex frontend to monitor and control flights live.
Slice-based CNN pipeline on RSNA Kaggle brain scans using modality-aware preprocessing and depthwise aggregation across slices. Achieved around 0.62 AUC for efficient aneurysm detection and localization.
Developed a neural network to predict peak oil production rates for Chevron. Used TensorFlow/Keras with hyperparameter tuning and achieved high accuracy through feature engineering and regularization techniques.
Created a flight route optimizer for HackRice 13 that finds carbon-efficient paths using a modified Dijkstra's algorithm. Integrates real-time flight data for environmentally conscious travel planning.
Built a predictive model for Chevron's 2023 Rice Datathon to forecast renewable energy investments by state. Used Linear, Lasso, and Ridge regression to optimize accuracy and identify collaboration opportunities.
Want to collaborate?
I'm always open to interesting projects and new challenges. Let's build something cool together!