Researchers at Stanford, UCLA, and SLAC have developed a deep learning surrogate model that accelerates nonlinear optics simulations by orders of magnitude compared to conventional methods.