Regular biography
Adit Radhakrishnan is an Assistant Professor of Applied Mathematics at the Massachusetts Institute of Technology, Department of Mathematics. He joined the department on July 1, 2025, and is also an associate member at the Broad Institute of MIT and Harvard. His research interests include understanding the mathematical principles underlying deep learning and applying these to develop new machine learning algorithms, particularly in the biomedical field. He has focused on how neural networks learn patterns from data and improving their feature learning capabilities. His work includes characterizing features learned by Large Language Models and applying research to understand patterns in genomics data related to cellular programs.
Scholar-generated biography
Adityanarayanan Radhakrishnan is a researcher specializing in Deep Learning, Machine Learning, Feature Learning, and Genomics. His work explores the mechanisms of feature learning in neural networks, including the role of backpropagation-free models and the dynamics of overparameterized networks. He investigates how neural networks implement associative memory and how cross-modal autoencoders can learn holistic representations of biological states. His research also extends to genomics, such as using machine learning for nuclear mechano-morphometric biomarkers in cancer diagnosis and causal network models for SARS-CoV-2 expression. Radhakrishnan's contributions span theoretical and applied aspects of AI, including quadratic models for neural network dynamics and transfer learning with kernel methods.