NdotLight is extending TRINIX, its text-based 3D CAD generation solution, from design automation into training data for Physical AI. Unlike generative 3D tools focused on surface appearance, NdotLight's proprietary 'moving CAD' technology includes components, joints, and physical properties, allowing one data source to support both design and robot learning.

While many generative 3D tools primarily create appearance-oriented meshes, TRINIX defines component-level structures and allows users to add or edit functional elements such as lids, handles, and joints. Because the generated CAD can behave according to physical rules, it supports both product design and robot simulation.

NdotLight has built pipelines that connect generated CAD with major 3D simulation platforms, including NVIDIA Omniverse. The company also contributes data to the K-Humanoid Alliance's robot foundation model and is preparing for demand that will expand from industrial robots to household humanoids.

TRINIX is optimized for each customer's terminology and design rules. Simple designs can be generated in about 40 seconds, while more complex products average roughly 90 seconds. NdotLight will continue improving its models through industrial deployment while strengthening protection for customer IP and security.