Architected Materials derive their effective mechanical behavior from the interplay between a base material and a mesostructural geometry. This enables the design of targeted mechanical responses and motivates inverse problem formulations, in which desired macroscopic properties are prescribed and the corresponding mesostructures are identified. In this context, data-driven inverse design approaches facilitate the navigation of the associated high-dimensional design spaces. In this work, we consider the inverse design of architected materials with a focus on spinodoid metamaterials [1]. Spinodoids are characterized by a small number of topological descriptors while exhibiting a wide range of anisotropic effective responses. Our proposed inverse design workflow relies on a neural network trained as a forward surrogate to establish a mapping between the spinodoid descriptors and the corresponding effective elastic properties. This surrogate model is subsequently evaluated within an optimization framework to determine the descriptors that produce a spinodoid with the targeted mechanical response. In addition to linear elasticity, we extend the inverse design framework to account for the onset of plasticity. Data-driven surrogate models of the initial yield surface are constructed from numerically generated stress-strain data using a kernel-based regression approach [2,3]. These yield surface surrogates are subsequently incorporated into the inverse design workflow, now allowing elastic and plastic response characteristics to be considered during the inverse design. REFERENCES [1] Kumar, S., Tan, S., Zheng, L. & Kochmann, D. M., Inverse-designed spinodoid metamaterials, npj Comput. Mater. 6, 73 (2020). [2] Fourier-Accelerated Nodal Solver (FANS): https://github.com/DataAnalyticsEngineering/FANS [3] Kunc, O., & Fritzen, F., Generation of energy-minimizing point sets on spheres and their application in mesh-free interpolation and differentiation, Advances in Computational Mathematics, (2019).