Wojciech Matusik Research Group

The group builds domain-specific design languages and solver-checked programs to generate correct-by-construction artifacts for digital manufacturing. They develop neural physics surrogates and learned simulators that preserve physical structure to accelerate or replace classical solvers while retaining invariants for trustworthiness. The group constructs closed-loop AI-for-discovery systems that propose candidates, run experiments on real instruments, and learn from outcomes to accelerate materials and device discovery. They implement differentiable simulators and inverse-design pipelines applied to cloth, multimaterial 3D printing, robotics, and composite jetting to enable practical inverse fabrication.