[RCAC Workshop] GeoAI on HPC: from single-task paradigm to multiple-task Geoscience Foundation Models
📅 Date: March 6th, 2026
⏰ Time: 1:30-4:30PM
💻 Location: VIRTUAL
🏫 Instructor: Xiao Liu
Please register using the link below to receive email reminders and the Microsoft Teams link; the “I’m interested” button does not provide access.
🔗 Register now: Click here to register
This workshop explores the evolving landscape of Geospatial Artificial Intelligence (GeoAI) on High-Performance Computing (HPC) environments.
We'll begin with the foundational concepts of GeoAI, introducing what typical GeoAI tasks are and how AI are applied to geospatial data like satellite imagery.
Next, we'll dive into the practical tools that enable this work, and set up working environment on HPC Clusters. We'll examine the TorchGeo library, which provides a comprehensive toolkit for training models on geospatial datasets within PyTorch. We'll also cover TerraTorch, a fine-tuning and benchmarking toolkit that extends capabilities for Geospatial Foundation Models.
Finally, we'll present several case studies to illustrate the transition from traditional, single-task GeoAI models—like a model trained to identify specific land cover types—to the latest advancements in Geoscience Foundation Models. These larger, more versatile models are capable of performing perform specific geoscience tasks with much less new data than would be required for a traditional model. We will also introduce three ways to perform these tasks with HPC, and mainly focus on two of them with online experiments: interactive job via OoD and batch job.