Machine Learning Potentials for Large-Scale Simulations

Abstract

This presentation showcases interdisciplinary research that bridges artificial intelligence (AI), high-performance computing (HPC), and materials science to advance semiconductor technology. This presentation is based on the Machine Learning Interatomic Potential (MLIP) paper featured in the NeurIPS 2023 D&B track. In this work, we introduce a novel semiconductor dataset (SAMD23) along with domain-specific evaluation metrics that are crucial for the development of generalizable MLIP models. These metrics are integrated into the framework, allowing for the effortless evaluation of MLIP models without the need for domain-specific knowledge. The goal of the presentation is to highlight real-world applications of machine learning in materials science, showcasing current trends and encouraging future innovation and interdisciplinary collaboration. This integration of AI, HPC, and materials science not only pushes the boundaries of technological advancement but also sets the stage for innovative work in semiconductor technology and beyond.

Bio

Image

Geonu Kim, a pioneering Staff Researcher at the Samsung Advanced Institute of Technology (SAIT), specializes in multidisciplinary research combining high-performance computing (HPC), artificial intelligence (AI), and materials science. Holding a Ph.D. in Materials Science and Engineering from Carnegie Mellon University and currently pursuing an M.S. in Computer Science at Georgia Tech, he leads semiconductor technology innovation by integrating AI with materials science.