Pratik Patil

Pratik Patil’s research develops mathematical foundations for modern machine learning and artificial intelligence. Because many of today’s most powerful models are extremely large and complex, behaving in surprising ways that classical statistical wisdom cannot easily explain, Patil builds theory to understand these behaviors and to provide scientists and engineers with practical tools for designing more efficient and reliable systems. Some of his past research has studied why very large models can sometimes perform well on new data despite overfitting; how combining many models into ensembles can improve accuracy; and how strategies like selective use of data or features can make models more efficient and stable. He has also developed new approaches to evaluate and compare models so practitioners can make confident choices about model size, training and tuning. Patil completed postdoctoral research at the University of California, Berkeley and received his Ph.D. from Carnegie Mellon University, master’s degrees from the University of Toronto and a bachelor’s degree from the Indian Institute of Technology.