Kowshik Thopalli is a Machine Learning Scientist and a postdoctoral researcher with the Machine Intelligence Group at Lawrence Livermore National Laboratory. His research focus lies in developing robust and adaptable machine learning models capable of performing consistently well under varying test conditions, including shifts in data distribution. Kowshik’s work has been published in prestigious machine learning conferences and Journals like ICML and ECCV, where he has presented his techniques for enhancing model resilience through methods such as domain adaptation, domain generalization, and test-time adjustment using geometric and meta-learning approaches. Kowshik’s expertise extends to combining diverse knowledge sources, such as guidance from subject matter experts and generative models, to boost the data efficiency, accuracy, and distribution shift resilience of machine learning models. Currently, he is collaborating closely with Physicists at the laboratory to create reliable surrogate models for complex simulations, with the goal of significantly improving computational efficiency.
Having obtained his Ph.D. in 2023 and his Master’s degree in 2018 from the Electrical and Computer Engineering department at Arizona State University, Kowshik was advised by Dr. Pavan Turaga and Dr. Jayaraman J. Thiagarajan. He firmly believes in the power of interdisciplinary collaboration as a critical factor for scientific success and actively engages with researchers from biomedical sciences, physics, and high-performance computing to tackle challenging scientific problems and transform them through the application of machine and deep learning techniques.