Newton & Networks: Physics in the Design Loop

Field notes

The site describes a workflow in which learned surrogate models stand in for expensive physics simulations during early design exploration, while full simulation remains relevant for validation. Its featured maritime example uses geometric deep learning to predict hull hydrodynamics, and the homepage reports a 10× speed improvement and a 100× speed-up over full simulation. Other visible work covers sensor-based anomaly detection and traceable retrieval or multi-agent systems for engineering documents.