Muhammad Talha

I am a dual-major Ph.D. candidate in Computer Science and Biosystems and Agricultural Engineering at Michigan State University, advised by Dr. Amirpouyan Nejadhashemi and Dr. Pang-Ning Tan. Most rivers on Earth have no gauge, so my research asks how models can learn from the places we do measure and carry that knowledge to the places we don’t. I work on foundation models for Earth systems, graph neural networks on river networks, spatiotemporal deep learning, and transferable basin representations for hydrologic prediction.

I am currently a Research Intern at Oak Ridge National Laboratory through the Graduate Research Opportunity (GRO) Program, working with Dr. Dan Lu. There I contribute to HydroORBIT, a Genesis Mission–funded multimodal foundation model for continental-scale hydrologic prediction, and build graph neural networks for distributed streamflow prediction.

My dissertation, Spatially Distributed Basin Representations for Hydrologic Prediction Across Space and Time, asks how to learn representations of river basins from distributed meteorological and landscape data that transfer to new places and new time periods.