Practical Materials AI for Improving Electrochemical Stability

The prospect of using AI for materials engineering has inspired a large volume of innovative academic work using unsupervised and supervised machine learning on materials data. However, making materials AI practical, particularly in industrial contexts, has proven elusive for reasons of insufficient data, a disconnect between simulation and real materials, and technical knowledge gaps between materials AI developers and materials science practicioners. In this talk, I will discuss case studies of efforts to develop materials AI tools at the Toyota Research Institute for the purpose of mitigating electrochemical degradation of materials. In that context, I will comment on what has been effective, adoptable by industrial researchers and engineers, and what has proven less useful. I will conclude by articulating a research strategy informed by these practical experiences, outlining our future efforts towards making Materials AI matter for materials scientists in the real world.