Events

CILVR Seminar: Adaptive learning for clinical and biomedical reasoning

Speaker: Shalmali Joshi (Columbia)

Location: 60 Fifth Avenue, Room 7th Floor Open Space
Videoconference link: https://nyu.zoom.us/j/98954373715

Date: Wednesday, October 14, 2026

Abstract: Large-scale, multimodal electronic health record (EHR) data has strong potential to generate insights about populations, subpopulations, and individual patients. Foundation models, machine learning models trained on large amounts of data to address many downstream tasks, have shown early promise toward this goal. However, fundamental challenges remain in ensuring that such models provide reliable inference about patient health. I will outline gaps that arise from the irregularly sampled, heterogeneous data typical of health and medicine. I will then show how recent developments in leveraging autoregressive training yield robust posterior predictives, which can, in turn, enable broader downstream prediction and adaptive data collection. I will close with insights from applying these methods to EHR data from Columbia University Irving Medical Center.

Bio: Shalmali Joshi is an Assistant Professor in the Department of Biomedical Informatics at Columbia University. She is a member of the Data Science Institute and is affiliated with the Department of Computer Science and the SNF Center for Precision Psychiatry and Mental Health. Before joining Columbia, she was a Postdoctoral Fellow at Harvard University and the Vector Institute. She received her PhD in Electrical and Computer Engineering from the University of Texas at Austin. Her research designs AI and ML systems that improve the causal reasoning and predictive capabilities of foundation models in biomedical science and informatics, with an emphasis on generalizability, reliability, and robustness. Her lab focuses on applications in psychosis related disorders in psychiatry and neurodevelopmental disorders, as well as clinical tasks such as radiology. She serves on the advisory boards of Machine Learning for Healthcare (MLHC) and the Association for Health, Learning, and Inference (AHLI).