PNNL

Abstract

In the fall of 2022, PNNL organized a ten-week long AI/ML Boot Camp to introduce staff to AI/ML methods for Earth science research. This was followed by a six-week intensive AI/ML Hackathon in early 2024, which teamed up domain scientists with data scientists to collectively apply AI/ML methods to domain-specific use cases. The goal of the two organized sessions was to develop multi-disciplinary partnerships and teach the participating domain scientists enough to continue to successfully implement AI/ML techniques in their future research. The boot camp consisted of weekly hybrid sessions taught by in-house experts. Each session first overviewed the fundamental concepts behind each ML method, followed by a hands-on tutorial/activity using PNNL’s research computing (RC) resources or Amazon Web Services (AWS). The hackathon saw the return of many of the bootcamp’s participants and instructors, but this time paired them together in teams to address a domain science question. The groups got together once a week to report progress and engage in active working meetings using Microsoft Teams breakout rooms with RC servers provided for developing and running models. Post hackathon, some of our domain scientists were able to independently continue to implement AI/ML techniques in their research and present their new findings at conferences, such as the HydroML Symposium hosted by PNNL in 2024 [HydroML].