Deep Generative AI and Liquid AI Predictive Models: Applications in Materials, Energy, and Health Sciences Rajiv Kalia, Yash Gandhi, Aiichiro Nakano and Priya Vashishta University of Southern California, Los Angeles, CA 90089

In this presentation, I will discuss the application of deep learning models designed to denoise electron microscopy data for the growth of nanoparticles. Additionally, I will describe the use of denoising diffusion models for generating novel peptide sequences and illustrate how liquid AI foundation models can accurately predict the free energies associated with the binding of these peptides to proteins. Furthermore, we will examine the forecasting of methane emissions from abandoned oil wells using liquid AI models. I will also demonstrate how agentic AI can effectively automate this diverse range of applications.