MetAI's MetGen Integrates NVIDIA Tech for SimReady Digital Twins
Taiwan-based startup MetAI, a participant in NVIDIA’s Inception program, is at the forefront of developing AI-powered simulation environments poised to redefine industrial and physical AI. Central to its mission of accelerating simulation-driven AI development, MetAI is integrating key NVIDIA Omniverse technologies, including NVIDIA USD Search microservices, into its proprietary digital twin generation platform, MetGen. This strategic integration enables the creation of detailed, three-dimensional “SimReady” environments directly from conventional two-dimensional CAD blueprints, beginning with the complex demands of the warehousing industry.
MetGen, MetAI’s sophisticated engine for generating SimReady environments, possesses the capability to transform standardized 2D CAD files and structured data into high-fidelity, fully operational digital twins. These virtual replicas are meticulously designed for a multitude of critical applications, including the rigorous training of AI models, comprehensive testing of robotics, precise refinement of automation logic, and the efficient generation of synthetic data. The recent integration of NVIDIA USD Search significantly streamlines this process, allowing users to instantly locate and retrieve standardized SimReady assets from connected libraries. This enhancement dramatically reduces the time required for environment construction and substantially boosts scalability, fostering a more agile and efficient development cycle.
Daniel Yu, Co-founder and CEO of MetAI, underscores the foundational nature of their work. “From layout blueprints to SimReady assets, we’re building the infrastructure needed to train the next generation of industrial and physical AI,” Yu states. He elaborates on the transformative impact of the NVIDIA USD Search integration, explaining, “By integrating NVIDIA USD Search, we’re enabling a closed-loop workflow—where users can generate assets with MetGen, build their own searchable libraries, and dynamically retrieve components to streamline simulation creation. This shortens setup cycles and unlocks scalable, intelligent environment generation.” This integrated approach signifies a leap forward in creating dynamic, searchable virtual asset libraries that can be leveraged across diverse simulation needs.
The effectiveness of this integration has already been validated and is currently undergoing pilot deployment. Key capabilities emerging from this collaboration include the seamless conversion of 2D CAD layout blocks into searchable metadata, facilitating the automatic retrieval of matching 3D assets from connected Universal Scene Description (USD) libraries. Furthermore, MetGen’s internal pipeline can generate any missing assets, ensuring a complete and accurate digital twin. Looking ahead, MetAI plans to integrate agentic AI tools, promising even more intelligent and interactive scene construction.
This development marks a crucial step towards realizing robust “Real-to-Sim” and “Sim-to-Real” workflows, initially focused on warehouse automation but with clear potential for expansion into other industrial domains. MetAI is actively exploring additional applications for this versatile workflow across sectors such as data center construction, advanced manufacturing lines, and advanced robotics training environments. As MetGen prepares for its beta release in August, the MetAI team remains committed to cultivating an extensible ecosystem of SimReady digital twins, specifically tailored to meet the evolving demands of industrial AI development.