NdotLight is connecting product design with robot-AI training data through TRINIX, which generates functional 3D CAD from text or images. Because TRINIX generates CAD that understands component structures and joints and can move realistically, it is clearly differentiated from appearance-based generative 3D and can address design automation and robot learning together.

Unlike models that represent appearance alone, TRINIX creates CAD with component hierarchies and articulated structures. Users can continue refining a result through natural-language instructions, such as changing the number of valve handles or editing how doors and drawers move.

The resulting models can simulate real actions such as folding glasses, moving scissors, or opening product doors. Multiple shapes and operating conditions can be derived from one source model, making the data useful for teaching robots how to manipulate objects.

Since joining NVIDIA Inception in 2022, NdotLight has advanced its integration with Omniverse and presented synthetic-data technology that creates defects directly at the 3D mesh level at GTC 2025. By linking this capability with web collaboration through Surfee, the company aims to reduce repetitive design work and improve industrial-AI data productivity.