Materials shape our everyday life, making moving within and between cities and communicating possible and facilitating our work duties. It is thanks to the development of new materials that the internet era started, bringing with it many changes in our lifestyles and habits. The discovery of new materials is at the heart of these changes, and despite the extensive work in this direction by many synthesis chemistry research groups worldwide and the continuous progress in reticular and supramolecular chemistry approaches, materials discovery is still majorly driven by serendipity. Indeed, new data methods have shown that if we classify materials according to their chemical / structural features, vast regions in the structure/chemistry space are still unexplored. Computational methods have been instrumental in creating databases of hypothetical materials, [1, 2, 3, 4] however, most if not all of those data-generated structures have not yet been synthesized in the laboratory, and thus remain in the theory realm. The reason for this conundrum[ 5] is that we lack knowledge on synthesis conditions (solvent, temperature, additives. . . ) / structure correlations. Recent works combining self-driven laboratories and machine learning techniques have strived to make a connection between synthesis conditions and structure.[6] On another, more fundamental-science oriented philosophy, other works have focused in finding molecular-level mechanistic correlations to better understand the physical chemistry underlying nucleation, crystallization and phase transition processes.[7-31] These studies, taken together with the body of experimental works devoted to uncovering mechanistic details of the self-assembly process, have shaped our current knowledge of materials discovery. The objective of this CECAM workshop is to summarize the computational developments in the field and discuss ways to tackle the fundamental challenges that we face, including how to bridge (and synergetically combine) simulations and experiments, how to tackle the multi-scale nature of the self-assembly problem and how to merge simulation and data science approaches to answer these questions.