Power complex and high-accuracy AI surrogate models for CFD and FEA with GeoTransolver model architecture. Now available in Rescale AI Physics as part of the NVIDIA PhysicsNeMo library, GeoTransolver is a transformer model architecture that uses Geometry-Aware Latent Embedding (GALE) attention to model relationships across complex 3D geometries and unstructured meshes. It is designed to generalize across changing operating conditions and learn transient, geometry-aware physical behavior, making it well-suited for surrogate models that need to predict aerodynamic, structural, or deformation behavior across a range of design variants.

Learn more about surrogate model development with Rescale AI Physics.