Teaching

• Teaching Assistant (full-time), SYDE - 572 Intro. to Pattern Recognition (Class size: 70) Winter 2024
• Teaching Assistant (full-time), MSCI - 436 Decision Support Systems (Class size: 87) Spring 2023
• Teaching Assistant (half-time), MSCI - 546 Advanced Machine Learning (Class size: 54) Winter 2023

Research Scientist – Lead Researcher

University of Waterloo, September 2022 - Present

• Led research in egocentric 3D pose estimation while supervising two undergraduate research assistants (URA), developing spatio-temporal video Transformer solution with PyTorch to address self-occlusions, coordinating data preparation and model variant testing across the team, resulting in a publication.
• Developed novel framework for efficient salient object detection with PyTorch Lightning and PyTorch Geometric using superpixel-based Vision Transformers and Graph Convolutional Networks, addressing fundamental challenges of heterogeneity with dynamic centroid positional embedding and Fourier Descriptors with a team from Microsoft Office Media Group while mentoring a URA through experimental validation and implementation.

Research Scientist – Research Contributor

University of Waterloo, April 2023 - August 2024

• Played a key role in the winning AutoICE Challenge team by optimizing PyTorch data pipeline efficiency, implementing memory parsing techniques that reduced model training time from 8 hours to 1 hour, and developing robust data augmentation strategies, contributing to multiple publications on sea ice classification using SAR and multi-sensor satellite imagery.
• Pioneered analysis of satellite channel importance through innovative application of DeepLIFT and feature correlation techniques with PyTorch, identifying optimal input combinations for sea ice classification, leading to a publication and improving model efficiency through informed channel selection.
• Supervised a URA in developing unsupervised methods for fly sperm motility analysis, proposing novel metrics to quantify motility without single-cell tracking. Guided research on synthetic dataset generation and ground truth validation to evaluate metric efficacy contributing to a publication.

Home