Project Grant 2505986

This $250,000 Project Grant award from the National Science Foundation (NSF) Mathematical and Physical Sciences program (CFDA 47.049) supports the development of a Unified Neural Operator (UNO) framework for trustworthy operator learning to create computationally efficient and robust surrogate models for scientific applications. The project aims to advance scientific machine learning capabilities, particularly in areas like predicting complex fluid flows and modeling plasma behavior in fusion reactors, by embedding operator learning techniques into a unified theoretical framework that combines the rigor of traditional numerical methods with the expressivity of modern AI. The award will fund a collaborative research effort led by the University of Utah to deliver certifiable and interpretable AI-driven surrogates that address key challenges in operator learning, such as lack of robustness guarantees and scalable training methods, to enable broader adoption of these techniques in high-stakes scientific applications. The project is aligned with presidential priorities in artificial intelligence and nuclear energy, and supports NSF's mission to advance scientific knowledge and secure the national defense.