AI in Science

People from around campus talk about how they use AI.  Two 20 minute talks. The audience is mostly physics, math, CS.   This week features Derek Paley (Aerospace) and Ramani Duraiswami (CS).

Derek Paley: Title: RoboScout DTC: Emergency Response for Automated Intelligent Assessment of Mass Casualty Incidents

Abstract: This talk will describe ongoing research in robotic triage at the University of Maryland under the DARPA Triage Challenge competition. UMD Team RoboScout aims to demonstrate a standoff sensing capability using COTS sensors placed on uncrewed air and ground mobile robotic platforms with AI-based casualty assessment algorithms that provide automated, real-time labeling of mass-casualty injuries in the field. The overall goal is to focus on assessing from a distance using non-contact, standoff signature acquisitions for the leading causes of preventable trauma death. The specific research objective is to apply tools from AI and perception, medical trauma and sensors, and robotics and autonomy to develop physiological signatures of severe injuries, data-driven models to detect them, and mobile platforms to collect the sensor data. UMD is a finalist in the upcoming competition in November 2026.

Bio: Derek A. Paley is the Willis H. Young Jr. Professor of Aerospace Engineering Education in the Department of Aerospace Engineering and the Institute for Systems Research at the University of Maryland, where he has been on the faculty since 2007. He served as Director of the Maryland Robotics Center (2019–2025) and the UMD Autonomous Micro Air Vehicle Team (2014–2024) and was a Sabbatical Fellow at The Johns Hopkins Applied Physics Laboratory in 2025-2026. Paley received the B.S. degree in Applied Physics from Yale University in 1997 and the Ph.D. degree in Mechanical and Aerospace Engineering from Princeton University in 2007. Paley’s research interests are in the area of dynamics and control, including AI and autonomy for national security and public safety. Paley is Fellow of the American Society of Mechanical Engineers, Associate Fellow of the American Institute of Aeronautics and Astronautics and Senior Member of the Institute of Electrical and Electronics Engineers. He served as an Associate Editor for AIAA Journal of Guidance, Control, and Dynamics, IEEE Transactions on Control of Network Systems, and IEEE Control Systems.

Talk 2 by Ramani Duraiswami

Abstract

My laboratory, the Physical Intelligence and Reality Lab, develops methods at the intersection of mathematical physics, sensing, scientific computing, and machine learning. In this seminar, I will describe two related research directions.

Differentiable physics through analytical gradients. Across science and engineering, forward models encode valuable domain knowledge. By making such models differentiable, we can integrate that knowledge directly into learning architectures and build more efficient pipelines for parameter estimation, inverse problems, and interpretable modeling in data-sparse settings. I will present examples spanning acoustics, signal processing, biology, and fluid mechanics.

Operator learning for partial differential equations. Numerical solvers for the operators governing physical systems have enabled the detailed exploration of individual problem instances. However, conventional solvers generally do not transfer computational effort from one solve to the next. Operator learning instead seeks to learn a surrogate for the underlying solution operator, enabling rapid inference for new inputs after training on a collection of solved instances. I will present our recent work on GAIA, a novel transformer-based architecture that addresses both forward and inverse problems in a single model.

Bio:

Ramani Duraiswami is a Professor of Computer Science at the University of Maryland, with appointments in Applied Mathematics and Scientific Computing, Electrical Engineering, Neuroscience and Cognitive Science, AIM, Robotics, and UMIACS. His research spans machine learning, scientific computing, and computational perception. His earlier work on fast multipole methods, GPU-computing and real-time spatial audio led to two company spin-outs. His lab's audio technology now powers millions of Oculus VR headsets, headphones, Android devices, and PCs via CEVA. He has published over 350 papers across computer science, acoustics, applied mathematics, and machine learning.