Speaker
Description
Modern machine learning (ML) is enabling increasingly precise studies of the Standard Model that were previously beyond experimental reach. Sensitive measurements across energy scales, such as the Higgs self-coupling and the top-Higgs Yukawa coupling, are vital to understanding the nature of the Standard Model and potentially uncovering evidence of new physics. I present a research program that applies cutting-edge AI algorithms across the experimental pipeline in order to drive sensitivity in top quark and Higgs boson measurements, demonstrated through two recent CMS analyses. I present the first study of Higgs production decaying to two W bosons with at least one hadronic W, which is enabled by transformer-based jet tagging and fine-tuning techniques. I also discuss the first measurement of four-top production in the all-hadronic final state, which is facilitated by ML-driven event discrimination and data-driven background estimation. I then talk about how ML is changing how we collect data in collider experiments, through the example of AXOL1TL, the first ML-based anomaly detection algorithm operating in the CMS hardware trigger. Finally, I discuss the development of next-generation jet reconstruction algorithms, including symmetry-preserving attention networks for multi-Higgs and multi-top hadronic final states. Together, these developments, spanning trigger hardware, offline reconstruction, and physics-level interpretation, illustrate how AI-driven innovation is deepening our understanding of the Standard Model while extending our reach beyond it.