Proteins fold into 3D structures that determine their function and underlie cellular processes. Recent advances in structure prediction have enabled models with experimental-level accuracy, opening the door to incorporating structure directly into machine learning pipelines for protein engineering. In this project, we will refine and extend our ML workflows to integrate structural features alongside sequence data to predict and engineer physicochemical properties such as thermal resistance and enzymatic activity. This shift marks a move from predictive modeling to generative design, requiring high-speed access to large protein datasets, embeddings, and model checkpoints. The work is part of a broader international effort to develop next-generation protein engineering tools with applications in biotechnology and medicine.