I am working on the development of a machine learning based augmentation for MOF-FF. The target is a molecular neural network potential that can accurately predict the force and energy difference between MOF-FF and a chosen electronic structure method. The model is therefore able to reproduce the potential energy surfaces with great precision, if the structure is within the frame of reference learned by the neural network, but is also able to extrapolate beyond the training set using the FF part of the model. The model will be restricted to intramolecular interactions and to limit the network size, MOF-FF‘s building block approach is utilized.