Accelerating Materials Discovery with HPC and AI

AI and ML have gained surging interest in materials discovery. Since the release of the first universal machine learning potential M3GNet in 2022, which demonstrated general crystal structure optimization and simulation capabilities, many models have emerged for zero-shot predictions across the periodic table. We present a framework combining cloud HPC with materials AI inference to explore vast chemical spaces efficiently. Emphasizing the importance of translating predictions to synthesis, we demonstrate how to bridge the gap between AI and real-world impact. Using Li battery electrolyte development as a case study, we show how this approach enables exploration of unexplored element combinations for new materials discovery. The talk will cover practical strategies for scaling up AI-guided materials discovery from prediction to experimental validation.