The introduction of machine learning into the computational chemistry field has been a milestone. ANAKIN-ME (Accurate NeurAl networK engINe for Molecular Energies), or ANI, is a transferable machine learning potential for organic molecules. It is successful at reproducing the chemical accuracy of its reference level of theory and bridges the gap between accuracy and computational cost. ANI is used through our TorchANI software, which is coded in Pytorch. This makes it flexible and easy to make modifications, which is important for extending the ANI methodology. My current work uses ANI-2x and aims at predicting atomic descriptors in order to test the conceptual DFT charge models.