seminar series

Reliability Along the Way: Progress and Failure Signals in AI Agents

AI systems are rapidly evolving from models that generate responses to agents that execute long-horizon workflows, interact with tools and environments, and take increasingly consequential actions. This shift changes the nature of the reliability problem.

Title TBA

Learning Non-Covalent Interactions at the Wavenumber Scale

Streamed on Oct 29, 2026

Helium on atomically thin substrates enables exotic quantum many-body physics, but predictions depend sensitively on the adsorption potential. For benzene and larger polycyclic aromatic hydrocarbons, we combine actively selected coupled-cluster calculations, multi-fidelity Gaussian process regression, and knowledge distillation from universal interatomic potentials to achieve single-wavenumber…

Machine Learning Methods for Molecular Design and Reaction Discovery

Streamed on Oct 15, 2026

This talk will discuss strategies we have developed for efficiently navigating chemical space, and the physical insights we gained during their design and deployment. The first part will focus on our work in developing methods that accelerate the prediction of properties of molecules and materials at multiple length scales.

Agentic Modular Chemical Language Model (mCLM-A) for Molecule Discovery

Streamed on Oct 1, 2026

Scientific discovery is one of humanity’s greatest engines of progress, yet the process itself remains painfully slow, expensive, and largely artisanal. Foundation models have transformed AI into a powerful engine for acquiring, organizing, and reasoning over human knowledge.

Chemistry-Accelerated Machine-Enabled Learning (CAMEL): From Theoretical Understanding and Design to Experimental Interpretation and Discovery

Streamed on Sep 17, 2026

Machine learning is transforming materials discovery, but its greatest impact may come not from applying general-purpose AI to materials data, but from constructing learning methods that embody chemical and materials-physics knowledge. I will illustrate this through interpretable representations; differentiable design using Effective Atom Theory (EAT), differentiable Equiformer fast EAT (DEFEAT)…

Digital Twins, AI-aided control and the Path to Scalable Quantum Hardware: From Sensing to Computing

Streamed on Sep 3, 2026

Quantum devices, whether sensors or computers, suffer from intrinsic device heterogeneity, environmental drift, and manufacturing variability that make them difficult to calibrate and control at scale. In this talk, I’ll present a unified digital twin framework for designing and controlling quantum hardware, grounded in work I led at MITRE’s Adaptive Quantum Sensing Program and recent scaling law…

Streamed on May 7, 2026

This talk examines how tools from machine learning and theoretical physics can be combined to better understand the internal representations of LLMs. Drawing inspiration from holographic duality, the speaker introduces a novel entropy-based measure to quantify how nonlocal learned features are in relation to input tokens—offering new insight into how information is structured and processed in these systems. Learn more.

Streamed live on Apr 23, 2026 

The many-electron problem has resisted solution for decades — but neural networks are changing that. In this AI-MI Seminar, Flatiron Institute’s Shiwei Zhang presents a physics-inspired computational approach that has already surpassed state-of-the-art methods for two-dimensional electron systems, revealing exotic quantum phases and opening new frontiers in materials science.  Learn more.

Streamed live on Apr 9, 2026

This talk challenges foundational assumptions in information theory — asking whether computationally bounded observers can extract more from data than classical frameworks like Shannon entropy or Kolmogorov complexity suggest. The speaker introduces epiplexity, a new formalization of information that captures what real-world learners can actually use, with practical implications for how we select, generate, and transform data to build better AI systems. Learn more.

Streamed live on Mar 26, 2026

Join us to explore how researchers are using AI to unlock chemical and biological knowledge hidden in scientific literature. The talk will highlight new approaches for automatically extracting data from papers and patents to build large-scale datasets that power molecule and protein design—advancing drug discovery, molecular property prediction, and AI-driven therapeutic development. Learn more.

Streamed live on Mar 12, 2026

Join us in exploring how researchers are combining molecular simulation, machine learning, and electron microscopy to better understand structural heterogeneity in complex chemical systems. This talk is presented by Erik Thiede, Assistant Professor of Chemistry at Cornell University. Learn more.

Streamed live on Feb 26, 2026

Join us in learning how a self-driving biomaterials lab powered by AI and automation is transforming material design to better integrate with biological systems, presented by Adam Gormley, Associate Professor of Biomedical Engineering at Rutgers University. Learn more.

Streamed live on Feb 12, 2026