This mini-symposium will present recent advances in machine learning (ML)-assisted multiscale design of materials [1], emphasizing computational mechanics and data-driven innovations across a wide range of applications. Early ML approaches accelerated materials discovery by screening large databases and uncovering complex links between atomic structures and properties. More recent developments in high-accuracy universal ML interatomic potentials and generative models have greatly improved predictive capabilities and sped up the discovery of materials across scales, from atoms to bulk components. ML now enables first-principles-level predictions of electronic, optical, and mechanical properties for large systems, opening new possibilities for structural, functional, and multifunctional materials. At the mesoscale, ML-assisted microstructure reconstruction and physics-informed models for solving partial differential equations are advancing the understanding of microstructure–property relationships critical for multiscale modeling. The integration of ML platforms into autonomous laboratories, combining quantum mechanical simulations, large language models, and experimental testing, is transforming the traditional approach to materials synthesis. Alongside showcasing these advances, the symposium will address ongoing challenges in scalability, interpretability, and coupling with multiphysics simulations. This symposium will invite contributions in: 1. ML interatomic potentials and surrogate models for accelerated evaluation of materials properties. 2. Generative and inverse design approaches for structural, functional, and multifunctional materials. 3. ML-assisted multiscale modeling for predicting and linking properties across atomic, microstructural, and macroscopic scales. 4. Autonomous and closed-loop materials discovery platforms, integrating simulation, optimization, and experimental validation.