How do I use it?

Finetune our pre-trained model on at least 30 simulations. If your simulations are CFD, now you can run CFD simulations on novel geometries in seconds. The same logic applies for any simulation type.

Mach-1.5 was finetuned on 2,500 car aerodynamics simulations.

How is this different from surrogate modeling?

GPT is to Natural Language Processing what Mach (our model) is to Surrogate Modeling. Our model has a more general understanding of how physics affects geometries than traditional reduced order models.

This means that Mach will model nonlinear physics on novel geometries more accurately.