Agenda

All talks and panel sessions will take place in University Hall in McNamara Alumni Center.

You can see the abstracts on the Abstract page.

7:30 Breakfast in The Commons

8:00 Welcome Remarks: Joe Konstan, CSE Associate Dean

Workshop Introduction: Vipin Kumar

8:30 GenAI Methods Session 1: How can science drive new foundations of GenAI?

Moderator: Aryan Deshwal, University of Minnesota

This panel explores how the unique demands of scientific discovery can inspire the next generation of AI methods. Scientific applications require capabilities that remain difficult for current generative AI systems, including reasoning beyond observed data, integrating physical and domain knowledge, operating under uncertainty, and supporting controllable scientific exploration. Rather than viewing science simply as an application area, the discussion will examine how scientific challenges can drive advances in foundation models, self-supervised learning, reasoning, and knowledge-guided learning, ultimately producing AI methods that benefit both science and mainstream AI.

Speakers

Anuj Karpatne, Virginia Tech, From AI for Science to the Science of AI: How Scientific Knowledge Can Guide the Next Generation of Generative AI

Amarda Shehu, George Mason University, What Autonomous Scientific Discovery Demands of GenAI Foundations

Charuleka Varadharajan, Lawrence Berkeley Laboratory, Using Earth System Challenges to Drive New Approaches for Generative AI

Alina Zare, University of Florida, The Strength to Say "I Don't Know": Why AI Needs Competency Awareness for Scientific DiscoverySpeakers and Panelists

Panelists

Erich Bloch, Indiana University

Anuj Karpatne, Virginia Tech

Amarda Shehu, George Mason University

Charuleka Varadharajan, Lawrence Berkeley Laboratory

Alina Zare, University of Florida

9:45 Break

10:00 GenAI Methods Session 2: What data and methodological advances are needed for GenAI-enabled science?

Moderator: Anuj Karpatne, Virginia Tech

As AI systems become increasingly capable of generating hypotheses, designing experiments, and acting as scientific collaborators, new methods are needed to evaluate their scientific value. This panel examines how to measure AI advantage in scientific discovery, moving beyond conventional machine learning benchmarks toward metrics that capture scientific insight, reproducibility, trustworthiness, and real-world impact. The discussion will explore how scientific evaluation can shape the future development of trustworthy AI systems.

Speakers and Panelists

Aryan Deshwal, University of Minnesota, AI-driven Adaptive Experimental Design for Accelerating Scientific Discovery

Nikunj Oza, NASA, GenAI-enabled Science for NASA

Manish Parashar, University of Utah, Addressing Data Challenges for AI-Enabled Science 

Ram Sriram, NIST, Measurement Science for AI-Enabled Scientific Discovery

11:30 Lightning Talks

Moderator: Vuk Mandic, University of Minnesota

Speakers: Shancong Mou, Derivative-Informed Training of Neural Operators On-the-Fly via Sketched Tangent Consistency
Mohammadali Maddahali, Distributed ML in adversary dominated environments
Ajay Kumar Gurumadaiah, Multi-Agent Video Prediction: Self-Correcting Conditional Frames for Dynamic Scene Forecasting
Wenkai Guan, AI's Hidden Thirst: A Predictive Decision-Support Tool for GenAI Data Center Water Circularity
Qianwen Wang, AI Data Scientists: Agentic Transformation from Data into Scientific Insights
Qianwen Wang co-presenting with Zhen Liu, GenAI-Enabled Representation of Quantum Chromo Dynamics

12:00 Lunch in The Commons and Thomas Swain Room

1:15 Lightning Talks

Moderator: Vuk Mandic, University of Minnesota

Speakers: Tian Cui, Ultrasensitive MEMS Sensors for Digital Health
Aryan Deshwal, Accelerating materials discovery with adaptively guided Generative AI
Ketson dos Santos, Advancing Machine Learning-Driven Polymer Design in the Electronic and Optical Domain with PolyGraphPy
Qizhi He, Generative AI for Inverse Modeling in Computational Mechanics and Materials Design
Peter Kang, AI-guided regime discovery for predictive subsurface systems
Pam Sooriyan, AI as a physics learning coach
Suo Yang, Graph GenAI for Interpretable Combustion Reaction Kinetics Discovery in Propulsion and Energy Systems

2:00 Break

2:15 Session: The Future of GenAI for Materials Research

Moderator: Chris Bartel, University of Minnesota

Materials discovery and design is of paramount importance to countless technologies including semiconductors, energy conversion and storage, aerospace, and biotechnology. Atomistic insights from first-principles calculations and molecular dynamics simulations have become central tools to predict the properties and performance of next-generation materials. In tandem with computational tools, new developments in automated and autonomous experimental platforms are coming on line to speed up the iteration from idea to observation. For both computational and experimental researchers, machine learning is playing an increasingly prominent role in accelerating materials science discoveries, from the development of interatomic potentials to the speed-up of ensemble predictions. Great strides have also been made with generative AI to propose new materials, produce hypotheses, and orchestrate workflows. This panel will discuss where the field is headed next and what challenges must be overcome to establish generative AI as an indispensable tool for materials scientists.

Speakers and Panelists

Vincenzo Lordi, LLNL, Frontiers of GenAI for Materials Science: Where are we and where are we headed?

Jin Qian, From Generative AI to Digital Twins: Building Trustworthy AI for Materials Discovery

Taylor Sparks, From Generating Materials to Designing Them: Reinforcement Learning for Scientific Foundation Models

Subramanian Sankaranarayanan, Materials Discovery Cloud – A multimodal discovery engine for microelectronics

4:00 Poster Session and Reception in The Commons and Thomas Swain Room

6:00 Dinner on your own

7:30 Breakfast

8:15 Recap from Day 1

8:30 Session: GenAI in Physics and Astrophysics

Moderators: Nadja Strobbe and Mike Wilking, University of Minnesota

This panel will examine the transformative potential of generative AI to advance research in physics and astrophysics. It will highlight applications ranging from modeling complex, non-linear interactions in particle physics detectors to searching tens of terabytes of daily telescope data for new astrophysical phenomena. The panel will also address current limitations of these approaches, including robust inference of fundamental physics parameters from data, the challenges of handling heterogeneous data at scale, and the difficulty of transferring models trained on well-observed regimes to data-sparse regimes.

Speakers 

Stella Offner, UT Austin, AI Reaches for the Stars

Nhan Tran, Fermi Lab, AI and particle physics: from collisions to collaborations

Panelists

Taylor Childers, Argonne National Laboratory

Michael Coughlin, University of Minnesota

Stella Offner

Nhan Tran

Kazuhiro Terao, SLAC National Accelerator Laboratory

10:00 Break

10:20 Session: GenAI and Healthcare

Moderator: Matt Johnson, University of Minnesota

Generative AI is rapidly transforming healthcare and biomedical research, enabling new approaches to clinical decision support, multimodal data integration, scientific discovery, patient engagement, and healthcare operations. This panel will bring together leaders from academia, industry, and federal funding agencies to discuss the current landscape of generative AI for health and medicine. Panelists will highlight emerging research directions, including large language models, multimodal foundation models, agentic AI systems, and domain-specialized models that integrate diverse health data. The discussion will also examine the scientific and technical challenges that must be addressed to enable safe, reliable, and equitable deployment in real-world healthcare settings. Panelists will share perspectives on the infrastructure, datasets, and interdisciplinary collaborations needed to accelerate innovation while maintaining trust and patient safety. The conversation will also explore funding priorities and opportunities that can help bridge the gap between methodological advances and clinical impact. The session will conclude with a forward-looking discussion on the future of generative AI in healthcare and the key scientific, technological, and translational challenges that the research community must address over the coming decade.

Speakers

Tarek Haddad, Medtronic, Foundation Models and Generative AI in Medical Devices: Opportunities and Validation Challenges

Rui Zhang, University of Minnesota, From Generative AI Models to Trustworthy Learning Health Systems

Hamid Tizhoosh, Mayo Clinic, Generative AI: Between Euphoria and Promise

Panelists

Lin Yee Chen, Lillehei Heart Institute, UMN Medical School

Tarek Haddad

Ben Teplitzky, Boston Scientific

Hamid Tizhoosh

Chris Yang, Drexel University

Rui Zhang

12:00 Lunch in The Commons and Thomas Swain Room

1:00 Session: Generative AI for Agriculture

Moderators: David Mulla and Shashi Shekhar

This panel will explore Generative AI (GenAI) research challenges and opportunities in agriculture, which is socially important for food, feed, fiber, and fuel. However, agriculture is facing formidable challenges due to growing demand, workforce shortage and aging, soil health risks, invasive diseases and increasing precipitation variability. To address these challenges, we ought to use all tools at our disposal including GenAI as highlighted in science policies such as AI Action Plan,  AI Innovation and Security, Adv. Regenerative Agriculture and Strengthening Farm Resilience, USDA’s Research Priorities and NIFA’s Role, and NASA’s Agriculture Mission. Agriculture challenges also reveal major limitations of current GenAI due to issues such as agricultural exceptionalism, spatial heterogeneity, limited economic margins, and complex interactions and feedback loops among sub-systems. These provide foundational GenAI research opportunities for balancing domain knowledge and inductive bias to improve spatial generalization, robustness, and interpretability towards a virtuous cycle of foundational and use-inspired GenAI research.

Speakers and Panelists

Rahul Ramachandran, NASA, From Foundation Models to Agentic Workflows: A NASA Strategy for Accelerated Scientific Discovery

Brian Stucky, USDA Agricultural Research Service, Open-source AI workflows and the future of agricultural research

Raju Vatsavai, North Carolina State, Embeddings as the Backbone of Modern GeoAI: From Foundation Models to Real-World Applications

Dimitris Zermas, Sentera, GenAI for Drone Imagery: Lessons from Production Agriculture

2:30 Break

2:45 Concluding Panel

Moderator: Manish Parashar

This concluding panel will synthesize the key insights that emerged across the foundational and use-inspired sessions and identify the most important opportunities at the intersection of AI and science. The discussion will examine the common scientific and technological challenges that span multiple disciplines, as well as the research advances, community infrastructure, and partnerships needed to fully realize the potential of generative AI for scientific discovery. Panelists will explore how academia, industry, national laboratories, and federal agencies can work together to accelerate progress through complementary expertise, shared infrastructure, coordinated investments, and open scientific collaboration. The panel will conclude by identifying a set of high-impact research directions and collaborative priorities that can shape the next decade of AI for science.

Panelists

Taylor Childers, Argonne

Vince Lordi, LLNL

Nikunj Oza, NASA

Brian Stucky, USDA Agricultural Research Service

Chris Yang, Drexel University (formerly NSF)

4:00 Closing Remarks and Reception in The Commons and Thomas Swain Room