2025–2026 Schmidt AI in Science Speaker Series Highlights
The Schmidt AI in Science Speaker Series returned to host 23 leading researchers at the University of Chicago’s Hyde Park campus. The series creates a recurring forum for the emerging community of AI in Science researchers to share methodological advances, discuss challenges, and forge new collaborations. Organized by the Eric and Wendy Schmidt AI in Science Fellows at the University of Chicago’s Data Science Institute, the 2025–2026 series convened experts from institutions from New York University and Stanford to UC Berkeley, Boston University, the Université de Montréal, in addition to presentations from UChicago faculty and fellows. Together, they demonstrated how AI is transforming scientific inquiry and discovery across fields as varied as cosmology, materials science, neuroscience, computational biology, and climate research.
Some Highlights From the Year
Jonathan Weare (NYU): Learning Predictions of Extreme Events from Time Series Data
Pritzker AI+Science Visiting Scholar Jonathan Weare (NYU) tackled a core problem in applied mathematics: predicting rare or extreme events from observational data. His examples spanned extreme climate events, engineering failures, and the conformational rearrangements of biomolecules. Weare presented new methods developed with his group and collaborators that have yielded surprisingly high accuracy even when working with datasets shorter than the timescale of the event.
Ethan Anderes (UC Davis): Denoising Diffusion Models for Extragalactic Foregrounds in CMB Lensing Analysis
Ethan Anderes (UC Davis) applied generative AI tools to one of cosmology’s most technically demanding problems. Extragalactic foregrounds introduce complex, non-Gaussian structures that bias small-scale analyses of the cosmic microwave background (CMB), particularly gravitational lensing reconstruction. Anderes presented a generative AI modeling approach using denoising diffusion probabilistic models trained on paired simulation maps, combined with a new inference technique called MUSE. The result is a promising framework for incorporating non-Gaussian random field modeling into CMB likelihood inference, a methodological advancement with direct implications for understanding the large-scale structure of the universe.
Aditi Krishnapriyan (UC Berkeley): Learning Physical Dynamics with Generative Machine Learning
Aditi Krishnapriyan (UC Berkeley) offered a panoramic view of how physics-inspired machine learning methods are reshaping scientific modeling. She addressed how neural networks can autonomously discover fundamental physical relationships from data, and how more flexible modeling choices enable the capture of physical dynamics across multiple scales. Drawn from applications in fluid mechanics, molecular dynamics, and quantum mechanics, her work represents a convergence of machine learning, numerical analysis, dynamical systems theory, and computational geometry that exemplifies the Schmidt Fellows’ commitment to methodological depth.
A Season of Disciplinary Breadth
The 2025–2026 Schmidt AI in Science Speaker Series reflected the full spectrum of AI-driven research at UChicago and beyond. The fall semester opened with Weare’s extreme-event predictions, continued with Milo Lin (UT Southwestern) on machine learning and molecular system; Ulugbek Kamilov (UW Madison) on computational imaging and inverse problems; Jason MacLean (UChicago) on AI and neural circuits, Ching-Yao Lai (Stanford) on ML for fluid dynamics and climate systems; and Pankaj Mehta (Boston University) on physics-inspired machine learning, a theme that would echo across the spring.
The spring semester brought equal richness. Catherine Pfister (UChicago) presented work at the intersection of ecology and AI-enabled data analysis. Hongliang Xin (Virginia Tech) and Chenhao Tan (UChicago) contributed perspectives from catalysis and natural language processing. Nicholas Jackson (UIUC) outlined a pragmatic roadmap for AI in polymer science, one prioritizing data quality and synthetic feasibility to move towards actionable discovery. Elena Villhauer showed how modern machine-learning techniques (e.g., deep sets, graph neural networks, and transformers) enable novel experimental searches for string theory. Laura Gagliardi and Rama Ranganathan showcased UChicago’s depth in computational chemistry and evolutionary biology. Rama Ranganathan, explored what statistical models of genome sequences can reveal about the fundamental design of proteins. Marco Biroli connected statistical physics and machine learning, while the April lineup—Madeleine Torcasso, Laurence Perreault-Levasseur (Université de Montréal), Ramon Nogueira, Emma Liu, and Jorge Jaramillo—wove together astrophysics, computational neuroscience, simulation-based machine learning, and neural modeling. John Kitchin (CMU) and Christy Landes (UIUC) rounded out May alongside Krishnapriyan, covering catalysis, scientific machine learning, and frontier optical microscopy.
Fellows at the Helm
As in previous years, the 2025–2026 Speaker Series was organized and driven by the Schmidt AI in Science Fellows—postdoctoral researchers embedded across the domain sciences who bring the series’ programming into direct contact with their own research communities. By selecting and hosting speakers in their fields, Fellows build relationships with senior researchers, deepen their scholarly networks, and bring the broader AI+Science community into conversation with their day-to-day work.
Cohort 2 Fellow Martin Falk said that “Organizing this year’s speaker series was definitely one of the highlights of my time as a Schmidt Fellow.” He added that “It was a really unique opportunity where I developed practical organizational management skills, and was also able to learn from speakers working across a wide range of fields, with many different perspectives on how AI can be combined with scientific knowledge.” The wide-ranging speaker roster—spanning ten UChicago faculty, external researchers from eight institutions, and international researchers—reflects the deliberate breadth the Fellows bring to the series.
Looking Ahead
The Schmidt AI in Science Speaker Series has become an enduring space where methodological innovation meets disciplinary depth, and where early-career researchers engage with the scientists shaping the future of their fields. With the 2026–2027 schedule to be announced in the fall, the series continues to serve as both a mirror of the field’s rapid evolution and a catalyst for the collaborations that will define the next chapter of AI-enabled discovery.
Many thanks to all 23 speakers, the Schmidt AI in Science Fellows for their outstanding work organizing the series, and Schmidt Sciences for making this ongoing program possible.
Written by Casey Keel.