Astronomy is entering an era where scientific discovery is increasingly limited not by the amount of data we collect, but by our ability to extract reliable physical information from it. We build physics-informed, uncertainty-aware, and trustworthy AI that embeds physical principles, symmetries, and probabilistic reasoning directly into modern machine learning, transforming AI from a black-box prediction tool into a rigorous framework for scientific inference. We develop methods for pixel-level analysis, probabilistic modeling, and interpretable scientific machine learning that maximize the information content of complex astronomical observations while providing robust uncertainty quantification and controlling systematic errors. Although our primary applications are in cosmology, the methodologies we develop are broadly applicable across astronomy and other data-intensive scientific disciplines.
Astro-AI Group
Department of Astronomy, University of Illinois at Urbana-Champaign