Rare events are ubiquitous in nature. Examples include chemical reactions, protein folding, nucleation, and self-assembly. All of these are characterized by the fact that they are very hard to model by straightforward dynamical particle-based simulations, precisely because they occur rarely (with respect to the fundamental timestep). To gain understanding in these processes, enhanced sampling methods exists that focus on the rare event itself. One of these, the transition path sampling method was developed to sample dynamical trajectories between predefined states, exponentially increasing the efficiency. Path sampling also allows revealing and identifying complex mechanisms, but to do so new algorithms are needed. In this project you will couple path sampling with machine learning in order to reveal the complex mechanism of a folding protein (model). Being general, the approach will be applicable to many other transitions as well. As such, this new approach will present a milestone in rare event simulations.