제목: Building a Community-Centric AI Ecosystem: Integrating Data and HPC at NSF NCAR​​

​​

​​​​​​​​​​​​​​​​​​연사: Dr. Thomas Hauser (NSF NCAR)

​

장소: 과학관 B102​호​​

Abstract:

​​The National Science Foundation’s National Center for Atmospheric Research (NSF NCAR) is building a community-centric AI Ecosystem for Earth System Science. This ecosystem tightly integrates a Research Data Commons with advanced high-performance computing (HPC) capabilities. NSF NCAR combines its extensive Earth science data repositories, the Derecho supercomputer, and scalable on-premise cloud infrastructure (“CIRRUS”). The AI ecosystem aims to automate the entire AI development lifecycle, encompassing data ingestion and preprocessing, model training, validation, and deployment.

​In my seminar, I will introduce the Community Research Earth Digital Intelligence Twin (CREDIT), an open platform for developing and utilizing data-driven weather and climate models. CREDIT lowers the barrier to entry for using AI in Earth System Science. Tasks that once required months of coding and supercomputer allocations can now be accomplished in days, allowing graduate students and researchers to focus on their science. CREDIT provides pre-configured and scientifically validated AI models, as well as custom pipelines for data ingestion, training, and forecasting. It integrates modern neural network architectures, such as WXFormer and FuXi, with physical constraints, including mass, moisture, and energy conservation.

​​During this seminar, I will share our progress in connecting the Data Commons with our high-performance computing system. I will also give an overview of CREDIT as a platform for AI-enabled Earth System Science. Finally, I will discuss ways to collaborate, including options such as co-advised projects, research exchanges, and other opportunities for students.​