• Utilizing Diffusion Model to reconstruct water molecule distribution within graphene ports and accelerate molecular dynamic simulations• Evaluating various pretraining methods for different invariant and equivariant neural networks (EGNN, Schnet, GNS, ForceNet, etc.) • Finetuning to predict potential energy and atomic force of inter-molecular systems, reducing force RMSE on some models by more than 50%• Constructing efficient Deep Learning pipeline for molecular dynamic research using PyTorch, PyG with visualization and containerization