Although physical simulation has become indispensable in scientific and engineering design and analysis, many real-time, many-query decision-making, and uncertainty quantification tasks remain computationally prohibitive with classical high-fidelity methods. Recent advances in artificial intelligence and machine learning have opened new frontiers in model order reduction (MOR), providing scalable, data-driven approaches that dramatically reduce computational costs while preserving accuracy and generalizability. These innovations are particularly powerful for parametrized systems governed by complex nonlinear partial differential equations, where they enable efficient surrogate modeling, rapid multi-query evaluations, and real-time decision-making. This minisymposium will highlight recent progress in MOR, with a special emphasis on the synergistic integration of physics-based techniques and AI/ML methods—including physics-informed learning, neural operators, manifold learning, and hybrid approaches that embed domain knowledge. Topics of interest include, but are not limited to: nonlinear approximation techniques; model reduction in high-dimensional parametric spaces; hyper-reduction for nonlinear problems; adaptive and error-controlled MOR strategies; structure-preserving MOR approaches; and machine-learning-enhanced surrogate modeling. Applications span optimization, feedback control, uncertainty quantification, and inverse problems in computational physics and engineering. By bringing together researchers working at the intersection of MOR, physics-based simulation, and machine learning, the minisymposium aims to foster cross-disciplinary dialogue, share emerging methodologies, and critically assess the role of AI in advancing the scope, scalability, and reliability of reduced-order modeling.