The rapid advancement of artificial intelligence has driven an unprecedented demand for computing performance, memory bandwidth, and energy efficiency. However, the continued slowdown of Moore’s law has made it increasingly difficult to meet these requirements through transistor scaling alone. Consequently, advanced packaging has emerged as one of the most important and effective approaches to sustaining performance growth.  However, the complex geometrical layouts and inherently multiscale nature of advanced packaging structures impose stringent requirements on electromagnetic simulation tools. These challenges motivate this dissertation to develop an efficient full-wave transient solver capable of modeling the electromagnetic behavior of advanced packaging structures with high accuracy while achieving significantly accelerated computational performance compared to existing methods. In this dissertation, the unconditionally stable Laguerre-FDTD method is adopted as the core algorithm due to its intrinsic advantages in modeling multiscale structures. The major contributions of this work are threefold. First, a planar non-orthogonal grid meshing technique is proposed to significantly enhance the geometric flexibility of the Laguerre–FDTD method for accurately modeling complex structures. Second, a novel matrix preconditioner is developed and combined with GPU acceleration to substantially improve computational efficiency. Third, an AI-assisted simulation framework is demonstrated for the first time within the Laguerre–FDTD context, showcasing its potential for new paradigm of electromagnetic analysis.