Large and diverse property data sets of relaxed molecules and crystals, resulting from compu- tationally demanding quantum calculations, have recently been used to train machine learning models of various energetic and electronic properties. We propose to advance these techniques to a level where they can also describe reaction profiles, i.e. reactive non-equilibrium processes which traditionally would require quantum chemistry treatment. The resulting quantum ma- chine learning (QML) models will provide reaction profiles for new reactants in real-time and with quantum accuracy. The overall goal is to develop a predictive computational tool which allows chemists to easily optimize reaction conditions, develop new catalysts, or even plan new synthetic pathways.