At Anumana, I implemented a cardiac electrophysiology research platform in Python and PyTorch for analyzing long, 12-lead ECG recordings. I also developed an ECG delineation model that achieved an F1 score of 0.91 and engineered post-processing that raised QRS onset/offset F1 scores to 0.98/0.99.
I trained a contrastive self-supervised encoder on more than 500K unlabeled ECGs, reducing labeled data requirements by 40% for downstream pacing classification. Earlier, as a Software Developer Intern at Sciforn, I developed full-stack features and reduced average API latency by 29%.
