ML platforms, data engineering

• Research in computational combinatorial optimization with applications in cooperative game theory and uncertainty quantification.• Developed an efficient Python package to empirically study and test bounds on the minimum composition norms of Choquet capacities on product spaces by solving randomized linear programs.• Supervised by David Saunders & Mario Ghossoub.

Recommendation systems & feature store

• Researched normalizing flows and generative quantile models for probabilistic time-series forecasting and anomaly detection - conducted experiments in PyTorch on synthetic distributions, image, and time-series datasets.• Published first-author research on a new family of deep quantile networks bridging optimal transport theory at ICLR 2022's Deep Generative Models for Highly Structured Data workshop.• Advised by Yaoliang Yu.

• Developed a self-supervised learning method on 3D point clouds using tools from persistent homology theory.• Trained state-of-the-art 3D point cloud neural networks on HPC GPU clusters leveraging Slurm scheduler.• Advised by Bo Wang.

• Built out data pipelines in the HPC clusters for transforming magnetic resonance images (MRI) from DICOM format into pre-processed PyTorch tensors.• Developed and trained a novel U-Net architecture on the processed images for automatic scar quantification on GPU clusters.• Advised by Bo Wang.

• Led research project in deep image segmentation models for automatic ventricular segmentation from MR images.• Advised by Bo Wang.