Energetic particles are ubiquitous in magnetically confined fusion plasmas. They contain a significant fraction of the plasma energy and are thus vital for the performance of fusion devices such as ITER. However, the presence of energetic particles and the fact that fusion plasmas are complex systems heated up to hundred million degrees result in instabilities that reduce the confinement of energetic particles. Understanding, predicting and controlling their transport and losses is of prime importance and constitutes our main goal. This is a high-dimensional multi-scale nonlinear problem, for which a complete description is so far unaffordable. Therefore, we propose a novel and inter-disciplinary approach to develop numerical tools based on Artificial Intelligence techniques applied to two lines of research: (1) derive data-driven reduced models for transport of energetic particles and (2) optimize the information extracted from HPC gyro-kinetic simulations and from experiments.