Learn geometric deep learning for proteins, ligands, and complexes. Represent structures as graphs and surfaces, train 2D/3D equivariant GNNs for pocket/interaction/affinity tasks, and ship reproducible evaluation reports with clear baselines and error bars.