3D AI technology startup Endotlight (CEO Jinyoung Park) has joined the Motif Technologies consortium and is launching a national project to build a unique AI foundation model for South Korea.
This consortium plans to gradually advance AI models, starting with a 300B parameter-level inferential language model (LLM) and moving on to visual language models (VLM) and visual language behavior models (VLA), and to drive the growth of the entire domestic AI industry ecosystem by releasing model weights, code, and computational optimization libraries as commercial open source.
EndotLight will be responsible for building the 3D data infrastructure for physical AI learning in this project and will play two key roles.
First, we're building an AI-based simulation-ready (Sim-Ready) 3D data generation pipeline. Using just text or image input, we generate precision 3D CAD data that can be manufactured and then convert it into a form ready for immediate application in a simulation environment, eliminating the time and cost issues inherent in traditional CAD-to-simulator conversion processes.
Second, the creation of large-scale synthetic data required for training VLA models for physical AI learning. This simulation-based approach provides the synthetic data necessary for training VLA models that integrate vision, language, and behavior, while reducing costs and ensuring infinite scalability and reproducibility. This infrastructure is considered essential for securing national AI robotics competitiveness.
Park Jin-young, CEO of Ndotlight, said, “The AI foundation model competition is expanding to physical AI, including understanding the physical world, and Sim-Ready 3D data infrastructure in the robotics and autonomous systems fields will determine national competitiveness.” He added, “Ndotlight will play a pivotal role in establishing the physical foundation of the domestic AI ecosystem and accelerating the AI transition across industries.”
- See more related articles
You must be logged in to post a comment.