Abstract
The convergence of artificial intelligence (AI), high-performance computing, and experimental science is rapidly transforming discovery in physics and engineering. This seminar will present the latest advances in real-time AI algorithms, with a focus on their deployment in large-scale experiments. Drawing on breakthroughs from the A3D3 Institute, I will showcase how deep learning, graph neural networks, and AI hardware co-design enable rapid data analysis and decision-making across high-energy physics, astrophysics, and neuroscience. Notably, I will discuss experiences from the NSF HDR ML Challenge, where participants tackled anomaly detection in domains such as gravitational waves and environmental monitoring, and highlight the integration of these techniques in competitions like the「2025 FPGA智慧運算與終端節點創意應用競賽」, which encourages innovative AI and FPGA solutions among university teams.
The talk will also address the challenges and opportunities of integrating AI with heterogeneous computing platforms such as FPGAs and GPUs as well as model compression for edge deployment and the development of foundation models for scientific data. Emphasis will be placed on collaborative interdisciplinary approaches that bridge physics and engineering with practical insights drawn from these international challenges and competitions. This seminar aims to inspire PhD students to engage with cutting-edge research at the intersection of science engineering and AI.