Spinodoids arise from spinodal decomposition, a thermodynamically driven phase separation process that produces complex bicontinuous topologies [1]. These structures are commonly modeled using stochastic spatial variations based on Gaussian Random Fields (GRFs), which capture key characteristics of the underlying phase transition process [2]. Owing to their large and tunable design space, spinodoid architectures offer significant potential for achieving multifunctional performance beyond that of traditional periodic architected materials. In this study, spinodoid structures are investigated with the goal of enhancing both mechanical and thermal performance. A deep learning surrogate model based on convolutional neural networks (CNNs) is developed to predict homogenized elastic and thermal properties directly as a function of topology. Finite element analyses (FEA) are employed to compute the mechanical and thermal properties required to construct the training dataset for the neural network. The trained model is subsequently coupled with a genetic algorithm (GA) to perform multi-objective optimization, targeting trade-offs between thermal conductivity and elastic properties. The proposed framework identifies optimized metal–metal composite spinodoid designs that exhibit improved multifunctional behavior. These results demonstrate the effectiveness of combining stochastic topology generation, deep learning, and evolutionary optimization for the computational design of advanced metallic metastructures with tailored mechanical and thermal properties. REFERENCES [1] Grant, Christopher P. Spinodal decomposition for the Cahn-Hilliard equation. Communications in Partial Differential Equations 18.3-4 pp. 453-490, 1993. [2] Kumar, S., Tan, S., Zheng, L., & Kochmann, D. Inverse-designed spinodoid metamaterials. npj Computational Materials, 6(1), 73, 2020.