Clarkson University
Machine Learning Researcher (REU) · Potsdam, NY
Clarkson is a research university in northern New York, and I spent the summer in Prof. Daqing Hou's group working on behavioral biometrics.
- Built a foundational model that identifies people from their mouse movement: a 2-layer BiLSTM in PyTorch that turns movement into 128-dimension embeddings, trained with semi-hard triplet loss.
- Trained one model across five public datasets (Balabit, ChaoShen, DFL, TWOS and Gmail) instead of one model per user, so it can check users it never saw in training.
- Presented the work at the Summer 2026 RAPS showcase and won Best Undergraduate Oral Presentation in Computational Science & Applied Mathematics II.
- Found that each dataset leaves a timing fingerprint from its logger's polling rate that the model could lean on, which became my Fall 2026 independent study at RIT.
PyTorchPythonMetric learningBiLSTMGPU clusters