❏ Applied neural networks with PyTorch to predict 3 phenotypes of Sprague Dawley rats based only on genetic variants (SNPs) and transcriptome wide associations
❏ Simulated rat genomes to create additional training data with generative adversarial networks (GANs), improving accuracy by 15%
❏ Utilized the High-Performance Computing cluster to leverage CUDA GPUs, interfacing with PyTorch Lightning
❏ Presented results in Columbia University’s Computer Science Research Colloquy
❏ Investigated deep learning's viability as a replacement for traditional polygenic risk metrics, benchmarking against 3 state-of-the-art classifiers - Under the supervision of Dr. Itsik Pe'er