This article shows the results of a research project developed by the University of Medellín, Kuepa company, Minciencias and the Government of Antioquia, which starts with the problem of designing virtual courses based on traditional pedagogical approaches, that they do not promote active learning and do not take into account student learning styles. For this, a project is developed in which new functionalities are added to a virtual learning platform to allow the detection of student learning styles, and the configuration of courses under ABP methodology (Problem Based Learning). To achieve this, artificial intelligence (AI) techniques are used. In the development of the article it is shown that the model used to determine learning styles was Kolb, this one was used as input to train a SOM (Self-Organizing Maps) neural network. In this way, the material and problems assignment are customized according to the student characteristics. This article shows the modules that make up the platform and the built-in neural network structure to provide intelligence to the system are shown.