Speaker
Description
Machine learning is now widely used in subatomic physics, yet most applications remain focused on task-specific models embedded in complex workflows involving data analysis, large-scale simulations, and inference. In this talk, I discuss a shift toward AI as scientific infrastructure, highlighting the emerging roles of foundation models, simulation surrogates, and AI agents in astrophysics and particle physics. Foundation models enable reusable representations that transfer across tasks and experimental conditions, surrogate models accelerate expensive simulations and inference loops, and AI agents provide a structured way to coordinate data access, simulations, analysis code, and provenance within large workflows. Rather than replacing expert reasoning, these approaches aim to improve scalability, reproducibility, and iteration speed in uncertainty-aware subatomic physics analyses.