Matthew McDermott PhD

Matthew McDermott received his PhD in Computer Science from MIT, studying representation learning for health and biomedicine with Professor Pete Szolovits. Subsequently, as a Berkowitz Postdoctoral Fellow at Harvard Medical School in Professor Zak Kohane’s lab, he builds high-capacity “foundation models” and other representation learning systems over EHR data. Some of his key prior works include Clinical BERT, one of the most widely used pre-trained clinical language models; Structure-inducing Pre-training, a framework for pre-training that enables incorporating domain-specific external knowledge with provable guarantees; and multiple software packages for performing machine learning at scale over structured EHR data, including the recent ESGPT package and MEDS framework. Matthew will join Columbia Department of Biomedical Informatics as an Assistant Professor in 2025, where he will continue his research in representation learning and foundation models over medical data.

Prior to his graduate studies, McDermott studied mathematics at Harvey Mudd College for his undergraduate degree, then worked as a software engineer at Google and as a co-founder of Guesstimate before his time at MIT. In addition to research, Matthew is also a leader in the machine learning for healthcare community at large, having served as a board member for the non-profit the Association for Health, Learning, and Inference (AHLI) since its inception and taking on numerous leadership roles for notable events such as the Machine Learning for Health (ML4H) Symposium, including general chair in 2021, and the Conference on Health, Inference, and Learning (CHIL), including as general chair in both 2024 and in the upcoming 2025 conference.