SE(3)-equivariant geometric deep learning model performing direct-shot blind docking: given a receptor PDB and a ligand, it predicts both the receptor binding location and the ligand's bound pose/orientation in a single forward pass.