Developed an end-to-end Geodite Graph Attention Transformer Network for atomic systems within the IBM/materials repository, focusing on improved modeling of atomic interactions and properties in materials science. The work involved designing modular encoders, decoders, and data loaders to process atomic features, laying the groundwork for training, inference, and evaluation pipelines. Leveraging Python and PyTorch, the implementation emphasized extensibility, enabling future experiments and model enhancements. The approach integrated deep learning and graph neural network techniques to address complex atomic system interactions, resulting in a robust baseline for further research and development in computational materials science without addressing bug fixes during the period.