AI in Nuclear Physics

Artificial intelligence is reshaping the landscape of nuclear physics, enabling new approaches to experiment design, data analysis, and real-time decision making at current and next-generation facilities. This talk will highlight recent advances driven by modern machine learning—deep neural networks, Bayesian inference, foundation models, and differentiable simulation—and their impact across the scientific workflow. I will discuss how AI is accelerating detector and experiment design through multi-objective optimization informed by high-fidelity simulations, reducing both the cost and turnaround time of exploring large design spaces. At the analysis level, AI-based reconstruction methods improve particle identification, track reconstruction, and calorimetric energy estimation, while uncertainty-aware models provide reliable, physics-informed predictions with quantified epistemic and aleatoric uncertainties. I will also explore the emergence of foundation models for nuclear physics, including tasks such as high-fidelity fast simulations of detector response and particle classification, which promise broad generalization across experiments. Finally, I will outline how these developments connect to the streaming-readout paradigm and autonomous, self-driving experimental workflows. Together, these advances position AI as a transformative enabler for the Electron Ion Collider (a primary example in this talk) and for other major nuclear physics initiatives worldwide.