"Particle physics has always been shaped by the computing technologies available to it. Each major advance has changed not only how we process data, but how we design experiments and make discoveries. In the 1950s, human scanners examined bubble-chamber photographs by hand. Today, the LHC experiments must filter events at rates of 40 million collisions per second, processing data streams of hundreds of terabytes per second under extreme latency constraints.
This talk discusses the evolution of computing in particle physics, focusing on how machine learning is reshaping every stage of the experimental pipeline. Neural networks now compress detector data on radiation-hard ASICs in nanoseconds, identify potentially new physics through anomaly detection in FPGA-based Level-1 triggers, and enable continuous readout and GPU-accelerated reconstruction. Looking ahead, we discuss how foundation models trained on particle physics data could transform simulation, reconstruction and analysis, and how agentic AI can shorten the path from data to scientific discovery.