Alumna accelerates battery discovery with AI

Speaking to a room full of materials science and engineering students in a recent graduate seminar, alumna Yumin Zhang, MSE’22 was drawn back to her own experience more than a decade ago. As a Ph.D. student, she worked with Venkat Viswanathan, who was a professor of mechanical engineering and held a courtesy appointment in materials science at the time. Zhang was drawn to his work on Density Functional Theory (DFT), a quantum mechanical modeling method used to accelerate battery development by predicting material properties. The interdisciplinary coordination among the engineering departments is what initially drew Zhang to Carnegie Mellon, and the connections she made spurred her career. 

“My (thesis) project was actually very similar to what I'm doing right now, designing liquid electrolytes,” Zhang said. “At the time, it was lithium batteries, but right now I'm working on all types of batteries.” 

Now Head of Research and Development at SES AI, Zhang has been responsible for developing their platform Molecular Universe, an end-to-end AI platform for battery materials discovery. Her work brings together expertise in materials science, machine learning, and software development.

During the seminar, she demonstrated the capabilities of the platform that she describes as something she “built from zero.” The tool combines domain knowledge with AI to accelerate battery discovery, with the goal of reducing battery research timelines from years to weeks using computational models, rather than using traditional trial and error experiments. The platform is a multi-functional resource for battery scientists, allowing them to ask, search, formulate, design and predict and has shown to find novel solutions compared to GPT or Gemini models. 

The impetus for developing the platform came as she was a member of an experimental team that would supply hundreds of molecules for screening each month, and company leaders encouraged her to use the data to build the tool, which has grown significantly in scope since its inception. Zhang credits the network that she started to build as a student at Carnegie Mellon as an impetus for her promotion from team member to team leader, now managing 40 people, including a team in Shanghai. 

“I became a manager because I had strong interpersonal skills and a good network in the battery field that I began building when I was a student at CMU,” she said, as she recalled opportunities that she had to share her work in various forums and conferences.  

Prior to her work with SEI, Zhang worked at ByteDance on their AI for Science team, where she was the only battery scientist, working with others with machine learning and software engineering backgrounds to study machine-learning force fields for battery liquid electrolytes. 

“I had gained in depth knowledge about liquid electrolytes and batteries at CMU, working on my Ph.D. thesis project and I also benefitted from ML courses I took.” 

Zhang thinks back to challenges that she faced as a student and now realizes the learning opportunities that they provided.  

“I learned how to address a research problem, ask critical questions, and design a project in a professional way,” she said. “Those skills have gone a long way.”