The MAMUT project aims at designing an open platform for solving recurrent and dynamic vehicle routing problems in urban environments. In order to address these urban transportation problems, we propose to combine operations research (OR) and artificial intelligence (AI). OR is able to propose fast and efficient resolution methods but these methods are often problem-specific and they do not take advantage of past solutions. AI, and especially machine learning, helps to define good solutions through the training of large historical data. The goal of the MAMUT project is to enhance the synergy between OR and AI for the resolution of vehicle routing problems. The objectives of the project are related to the scientific challenges, namely: (1) Identify, learn, and explain the characteristics of a class of urban logistics problems using machine learning, (2) Propose a large-scale hybrid RO/IA solver based on explainable AI rather than black-box ML methods, (3) Provide the scientific and industrial community with the problems, instances, algorithms developed through a collaborative platform.