Quantum mechanical accuracy is required to simulate hybrid organic-inorganic interfaces. Here we focus first on the structures and properties of molecules on surfaces. A problem we face is that accurate calculations are costly and extensive sampling is prohibitive, so studies into molecular assembly and surface-supported processes like diffusion are guided by human intuition. To promote unbiased studies into molecular surface structures and phenomena, we have combined atomistic simulations with Bayesian optimisation - an artificial intelligence (AI) technique designed for complicated optimisation tasks . We demonstrate how the AI was adapted to learn surface and property landscapes of molecules on surface with minimal computational sampling , delivering most stable surface structures with favorable designer properties. Energy landscapes can be further data-mined for low energy paths and associated trajectories to reveal the atomistic mechanisms behind key processes. We showcase the capability of AI to infer complex properties on several examples of atomic and molecular surface adsorbates (see Figure 1).
 M. U. Gutmann, J. Corander, J. Mach. Learn. Res. 17, 1 (2016).
 M.Todorović, M. U. Gutmann, J. Corander and P. Rinke, arXiv:1708.09274 (2017).