The simulation of dynamical processes in chemistry is still a substantial challenge because of the high complexity of realistic structural models and because of the need to reach statistically converged results. In spite of the availability of high performance supercomputers the use of ab initio molecular dynamics simulations is still prohibitively expensive for many interesting problems. In recent years, machine learning potentials have become promising new tools to transfer the accuracy of first principles methods to larger time and length scales. In this project we seek candidates to develop and apply high-dimensional neural network potentials to solve chemical problems in different fields like heterogeneous catalysis, vibrational spectroscopy and chemistry in solution.