We present a framework for an efficient and accurate multiscale analysis of heterogeneous materials. The framework addresses two major existing challenges: a) very high computational costs of directly coupled multiscale analysis, and b) introduction of additional human bias and knowledge uncertainties due to any macroscopic surrogate modeling of multiscale data. To tackle these challenges, we use a data-driven framework [1] which operates directly on multiscale data efficiently generated by FFT-based solvers [2]. In particular, we consider a 3-dimensional two-phase microstructure with elasto-plastic material behavior. A strategy for data acquisition as well as related questions regarding the solver scheme using FEM libraries are discussed. We demonstrate the framework by a selection of numerical examples. [1] Prume, E., Reese, S., Ortiz, M. (2023). Model-free Data-Driven inference in computational mechanics. Computer Methods in Applied Mechanics and Engineering, 403, 115704. [2] Moulinec, H., Suquet, P. (1998). A numerical method for computing the overall response of nonlinear composites with complex microstructure. Computer Methods in Applied Mechanics and Engineering, 157(1–2), 69–94.