Christos Thrampoulidis

Christos Thrampoulidis is an Assistant Professor in the Department of Electrical and Computer Engineering at the University of British Columbia (UBC), Vancouver. Previously, he held positions at the University of California, Santa Barbara (UCSB). His research focuses on machine learning theory, deep learning principles, mathematical optimization, and statistical signal processing. He leads the Machine Intelligence and Learning Design (MILD) group at UBC, emphasizing foundational studies of large language models, neural collapse, and high-dimensional data analysis.

Education: PhD and MS in Electrical Engineering from the California Institute of Technology (Caltech), with a minor in Applied and Computational Mathematics. Undergraduate degree (Diploma) in Electrical and Computer Engineering from the University of Patras, Greece.

Research interests include the implicit geometry of next-token prediction, optimization bias in neural networks, and robust generalization in overparameterized models. Key contributions include work on adversarial training, neural collapse phenomena, and safe reinforcement learning. His group actively explores theoretical frameworks for language models and develops methodologies to analyze their mechanics and limitations.

• Awards: NSERC Discovery Grant (2021).

Teaching includes courses on optimization and signals & systems at UBC. He has advised numerous graduate students (PhD and MSc) and undergraduates, focusing on rigorous mathematical approaches to machine learning problems. Active collaborations span academia and industry, with publications in top venues such as NeurIPS, ICASSP, and AAAI.

Current projects include understanding language model geometry, decentralized learning, and mitigating imbalances in training data. The group prioritizes equity and diversity, welcoming students with strong mathematical backgrounds and curiosity about foundational ML questions.