PhysIQ is a full-stack platform that uses Graph Neural Networks to predict fluid dynamics and cloth simulations 10-100x faster than traditional finite-element solvers. It includes a built-in inverse design engine using CVAEs and gradient-descent optimization to automatically find engineering shapes that hit specific performance targets.