BeyondMath reports that its Physics Factory platform autonomously developed a physics AI model using the DrivAerML automotive aerodynamics dataset. Given a plain-language objective, its AI agent crews analysed the data, composed model architectures, trained and evaluated experiments, and selected a final model with limited human involvement. The model ranked first for surface pressure, volume pressure and drag among published benchmark results, while another model remained ahead on velocity and lift. The approach demonstrates how AI agents could help engineering teams develop physics models from their own simulation data with limited specialist AI involvement.
Image courtesy: BeyondMath
