Recent developments in scientific machine learning and AI are reshaping the landscape of predictive computational modeling by enabling the fusion of heterogeneous data sources, including experimental measurements, imaging, and high-fidelity simulations. These integrative approaches are critical for advancing model credibility, informing high-consequence decision-making processes in complex physical systems, such as digital twins, optimal design under uncertainty, and experimental design. This mini-symposium will highlight novel research at the intersection of uncertainty quantification and machine learning, with an emphasis on scalable, trustworthy, and reliable predictive modeling frameworks. It will provide a forum for discussing foundational advances, algorithmic innovations, and application-driven studies that leverage uncertainty-aware machine learning modeling. Contributions addressing theoretical, computational, and software aspects are welcome. Topics of interest include (but are not limited to): • Uncertainty quantification (UQ) techniques in deep learning models • Model selection and credibility assessment • Optimal design, control, and decision-making under uncertainty • Operator learning and neural surrogates • Physics-constrained generative models and latent representations • Scalable UQ and inference algorithms for high-dimensional systems • Discovery of governing equations and interpretable model structures from data • Software frameworks for large-scale Bayesian inference and UQ • Applications in materials, biomedicine, climate, aerospace, and energy systems