In this talk, I will discuss how to apply graph convolutional neural networks to quantum chemistry and operational research. The same high-level paradigm can be applied to generate new molecules with optimized chemical properties and to solve the Travelling Salesman Problem. The proposed approach consists of two steps. First, a graph ConvNet is used to auto-encode molecules and estimate TSP solutions in one-shot. Second, beam search is applied to the output of neural networks to produce a valid chemical or combinatorial solution. Numerical experiments demonstrate the performances of this learning system.