Katabi Lab

The Katabi Lab at MIT is interested in the intersection of machine learning and physiological signals, focused on discovering biomarkers for neurological and inflammatory diseases as well as new machine learning techniques for this unique intersection of health and AI. On the health side, one focus area for our lab is sleep. Many neurological and inflammatory diseases are known to impact sleep, but their exploration in this domain has been limited due to the size and complexity of physiological signals. The approach is to use state-of-the-art machine learning models to assist in understanding these complex signals, and developing new tools to predict, diagnose or monitor diseases that other otherwise difficult or impossible to assess in the current standard of care. We publish our work in major medical journals including Nature Medicine, Science Translational, SLEEP, etc. On the machine learning side, our focus in on exploring self-supervision in the Computer Vision and Natural Language Processing domains. Work spans projects in representation learning and generative AI, especially focused on imbalanced data. We publish our work in major Machine Learning conferences including NeurIPS, ICML, ECCV, etc.