My research focus lies at the interplay between control theory and machine learning. On one side, control theory is a powerful theoretical tool to study the behavior of closed loop systems but it can be limited by the prediction capabilities of inaccurate dynamical models. On the other side, machine learning provides a vast set of techniques to continuously improve the accuracy of dynamical models but it lacks stability and safety guarantees that are crucial for safety-critical dynamical systems. In my research I try to make use of both techniques to build provably safe learning agents.