Engineering materials are inherently heterogeneous at micro- and meso-scales, often exhibiting variability due to manufacturing imperfections or natural processes. Capturing the mechanical response of such materials at the macroscopic scale requires a modeling framework that accounts for both aleatory (random) and epistemic (knowledge-based) uncertainties. This minisymposium explores modern approaches to multiscale material modeling that integrate stochastic methods and machine learning to represent and propagate uncertainty across scales. A central focus lies in learning effective macroscopic constitutive laws from microstructural data and understanding how uncertainties at lower scales affect system-level behavior—particularly in nonlinear regimes. We invite contributions in the following areas: • Stochastic modeling of heterogeneous and anisotropic materials • Stochastic and classical multiscale methods • Scientific machine learning for uncertainty-aware multiscale modeling • Uncertainty quantification in nonlinear material behavior • Model order reduction techniques for multiscale problems