Graduate Student Seminar
March 20, 2026
10:00 a.m. ET
CUC McConomy Auditorium
Accelerating Battery Materials Discovery with AI- and Physics-Powered Molecular Universe Platform
Advanced metal-ion (graphite or silicon anodes) and metal batteries are multi-component systems operating under a wide range of electrochemical conditions. Among these components, the electrolyte—present at every interface—faces the most stringent constraints and is often the last, most stubborn variable to optimize. Despite decades of research, systematic exploration has been limited to only a few hundred solvents, a small number of silicon-anode additives, and fewer than a dozen commonly used salt anions, most of which were adopted from general chemical catalogs rather than designed specifically for batteries. Yet under practical design rules, the viable space still spans ~10^11 neutral organics and ~10^9 anions, demanding a platform that can map and navigate this landscape efficiently. More importantly, molecule design is only the first step—the ultimate challenge is to bridge molecular properties to formulation behavior and cell-level performance.
In this seminar, I will present Molecular Universe (MU), the first comprehensive, end-to-end artificial intelligence (AI) discovery platform, unifying: (i) Map/Search—large-scale molecular databases covering solvents, salts, additives, and diluents, together with solid-state electrolyte property databases; (ii) Ask—domain-tuned large language model agents (including a multi-agent “Deep Space” mode) that synthesize insights from more than 60,000 publications, patents, and proprietary datasets to propose constraint-aware solutions; (iii) Formulate—physics-grounded simulations based on polarizable force-field molecular dynamics (MD) to predict formulation-level properties such as viscosity, solubility, miscibility, and conductivity; and (iv) Predict—cell-level machine-learning models that link molecular and formulation inputs, together with early cycling data, to performance endpoints such as cycle life. I will demonstrate how MU brings expert-level battery knowledge into an integrated AI-driven workflow at users’ fingertips and discuss the future outlook for “AI4Batteries.”
Molecular Universe is freely accessible to students at https://molecular-universe.com/about
Yumin Zhang
Senior scientist and technical project manager, SES AI
At SES AI, Zhang leads the development of Molecular Universe, an end-to-end artificial intelligence platform for battery materials discovery that integrates large-scale computational and experimental data, scientific literature, and machine-learning and large language models. Prior to joining SES AI, she was a research scientist on the ByteDance AI for Science team, focusing on machine-learning force fields for battery liquid electrolytes. She co-developed BAMBOO machine-learning force-field framework and is a co–first author of “A predictive machine learning force-field framework for liquid electrolyte development,”published in Nature Machine Intelligence (2025).¹ This work was adopted early in BYD’s (world’s largest EV maker) R&D process and is a key contributing factor to BYD’s Megawatt Flash Charging Battery (400km range in 5 minutes fast charging). Yumin received her Ph.D. in Materials Science and Engineering from Carnegie Mellon University, where her research focused on computational design of battery electrolytes. Her work lies at the intersection of molecular simulation, machine learning, and energy storage.
-
Gong, Sheng, Yumin Zhang, Zhenliang Mu, Zhichen Pu, Hongyi Wang, Xu Han, Zhiao Yu et al. "A
predictive machine learning force-field framework for liquid electrolyte development." Nature Machine
Intelligence (2025): 1-10.
Upcoming Events
-
February 20 2026
10:00 AM ET
Materials Science and Engineering
High-Fidelity Atomistic Simulations of Chemistry-Microstructure Interactions in Metals, presented by Rodrigo Freitas, Massachusetts Institute of Technology
CUC McConomy Auditorium
-
February 20 2026
5:30 PM - 9:30 PM ET
Highmark Center for Health, Wellness and Athletics (100 Tech St, Pittsburgh, PA 15213)
-
February 24 2026
12:00 PM ET
Materials Science and Engineering
"Modeling the Nonequilibrium Dynamics and Rheology of Associative Polymer Networks," presented by Songyue Liu
6142 Scott Hall
-
February 25 2026
2:00 PM ET
Faculty Insights with Mario Berges
Please join us for CMU Engineering's virtual program, “Faculty Insights: A 20 Minute Briefing.” In this series, faculty will share insights into their research, its impact, and provide perspective for the future of the field.
Virtual, link provided to registrants
-
February 27 2026
10:00 AM ET
Materials Science and Engineering
AI meets Peer Review: The Good, The Bad, and The Ugly, presented by Nihar Shah, Carnegie Mellon University
CUC McConomy Auditorium
-
March 13 2026
10:00 AM ET
Materials Science and Engineering
presented by Rachel Goldman, University of Michigan
CUC McConomy Auditorium