I’m an undergraduate at the University of Toronto (Computer Science and Statistics) working in computational medicine and biomedical imaging. I’m interested in building machine learning systems that hold up in the real world—across hospitals, across messy data, and under practical constraints. Recently, my work has focused on large-scale ECG representation learning and on automation for serial-section electron microscopy to make high-throughput imaging more reliable and less time-consuming
Developing large-scale ECG representation learning to improve generalization across hospitals. The focus is on careful evaluation, avoiding leakage, and measuring transfer gains on clinically relevant downstream tasks—especially where labels are limited or inconsistent.
Designed a hybrid strategy that uses fast low-resolution scans plus learning-based enhancement and targeted re-imaging only where needed—so high-resolution time is spent where it actually changes downstream results. Poster (PDF)
Built an automation pipeline that reduces tedious microscope interaction during long acquisition runs. The system helps keep imaging consistent across many sections and saves researchers substantial time by streamlining repeated selection and monitoring steps.
