Machine Learning for Oregon Wildlife Counting and Detection from Camera Trap Images
Project completed in partnership with Oregon Department of Transportation (ODOT). Paper available on request. 
Projects
A Multi-Planar Graph Visualization of Transformer Multi-Head Attention
Worked on creating a novel, interactive visualization for Tranformer neural network multi-head attention. Implemented as a wrapper for the Pytorch Transformer implementation. 
Visualizing the Loss Landscape of Neural Nets
Trained CNN models on the CIFAR-10 dataset and visualized the weights during backpropagation as a scalar field using dimensionality reduction. Inspired by the 2018 NeurIPS paper from Hao Li et al. 
AlphaGo Zero Lite
Developed a light-weight implementation of AlphaGo Zero that can be trained and run on a personal laptop to play board games through a command-line interface. 
SMT Solvers for Neural Network Training
Developed an algorithm to train simple feedforward neural networks with ReLU activations using SMT solvers. 