Miles Cranmer

Miles Cranmer is Assistant Professor in Data Intensive Science at the University of Cambridge, holding joint appointments in the Department of Applied Mathematics & Theoretical Physics (DAMTP) and the Institute of Astronomy (IoA). His research addresses the fundamental challenge of out-of-distribution generalization in machine learning for science – developing AI that can discover patterns that hold far beyond their training data. Dr. Cranmer created PySR, the standard open-source framework for symbolic regression, which he used to pioneer techniques for distilling mathematical models from deep learning systems. His research motivation also drives his other research efforts, which include work on the “PolymathicAI” collaboration to build industry-scale multi-discipline foundation models for numerical data. Through PolymathicAI, Dr. Cranmer and his research group build, interpret, and apply large-scale AI models to unlock new insights across diverse scientific frontiers: from cosmology and fluid dynamics to the mathematical underpinnings of AI itself. Before travelling across the pond, he earned a Ph.D. in Astrophysical Sciences from Princeton University and a B.Sc. in Physics from McGill University.

AI2050 Project

Cranmer is building technology that unlocks scientific discoveries hidden inside AI systems. Today’s AI models often outperform our best scientific theories in real-world systems but can’t explain why. His framework reveals the mathematical laws that AI has discovered but cannot express. By combining AI interpretation with symbolic mathematics, he transforms AI from a black box into a scientific collaborator. This could accelerate discovery across all sciences, giving scientists powerful new ways to understand nature.