Supercomputers have provided researchers with an unprecedented level of computational power. However, "power without grip is useless": the mere availability of thousands of processors must be accompanied by significant advances in software development and HPC techniques in order to effectively tackle the most complex simulation problems in computational physics and engineering. In particular, in application domains such as industry, energy, the environment, and biomechanics, the simulation of complex and often coupled fluid–solid phenomena remains a major challenge, demanding the full utilization of available computational resources. One distinctive feature of such large-scale simulations is their ability to generate massive volumes of data. In recent years, artificial intelligence—especially neural networks—has been employed to enhance simulations through surrogate and reduced-order modeling, to reconstruct flow fields from incomplete data, and to explore high-dimensional design spaces using reinforcement learning and other advanced methods. The objective of this Mini-Symposium is to foster discussion and exchange on current challenges and future directions in HPC simulations and AI techniques, with a particular focus on real-world applications spanning a broad range of fields, including biomechanics, automotive engineering, aerospace, pharmacology, energy, and environmental science. Relevant topics include algorithms, simulation strategies, and programming techniques for complex multiphysics simulations requiring massive HPC environments. Contributions addressing related aspects such as performance optimization, robustness analysis, and integration with pre- and post-processing tools (e.g., CAD, mesh generation, visualization) are also highly welcome.