Pranav Singh

Master's

Computer Science

New York University

Research Assistant I

• Architected novel deep learning pipelines for medical image segmentation and classification on a single GPU node, which improves upon previous benchmarks by 26% for disease classification and 5% for image segmentation. • Constructed post-processing architecture for segmentation which improves upon traditional segmentation architectures by 3% at no extra space cost and minimal computational cost. • Developed resource-efficient self-supervised learning approach which is twice as computationally efficient and improves upon existing state-of-the-art approaches by 8.5%, 7.3% and 11.5% for 1%, 10% and 100% label fractions respectively