The long-term vision and deliberate pace of E3SM capability development may sometimes be at odds with immediate modeling needs--for example, to support questions of climate impacts. ML/AI models can provide a fast, convenient solution where other modeling capabilities may not yet exist or be too expensive to deploy. We have found that the development of ML/AI models offers benefits beyond just "providing answers." Here, we demonstrate three examples where data science and machine learning have proven synergies across the InteRFACE project.

In one case, an AI model of Alaskan stream temperatures allowed a rapid assessment of impacts to Alaskan fisheries while collating observational data useful for validation. In the second case, AI modeling of river ice breakup led to data synthesis and predictions useful for developing ice-prediction capabilities in MOSART, while providing a reasonable model that predicts how climate may impact river ice and transportation in Alaska. The third example details CUSP--the CommUnity near-Surface Permafrost dataset. A goal of CUSP is to help unify the permafrost community while providing an off-the-shelf dataset for building ML models of permafrost presence and active layer thickness. The dataset itself is useful for RGMA efforts, and many Arctic MSD questions require an understanding of permafrost dynamics that CUSP + ML provides.

Finally, a brief description of a novel, AI-ready, hydro data platform called VotE will be given. VotE has enabled benchmarking CMIP6 + E3SM streamflows, assessing changing Arctic hydrology, and building an Arctic-CAMELS (paired meteorology and streamflow) dataset in support of the InteRFACE project.