San Jose, California, March 16, 2026 /PRNewswire/ — At NVIDIA GTC 2026, Linker Vision, a leading AI platform company and pioneer in physical and inference AI, will showcase its video inference AI platform designed to power intelligent operations across smart cities and smart spaces. Linker Vision, which has deployments in Taiwan and Vietnam, is working with Voxelmaps and Inchor to develop digital twin traffic simulation capabilities to test the potential of city management agents that can perceive, simulate, reason, and act in the city of San Jose.
The platform enables contextual understanding of live video streams to support decision-making across situation assessment, impact analysis, traffic operations, industrial monitoring, infrastructure management, and public safety. Linker Vision leverages the NVIDIA Metropolis Blueprint for Video Search and Summarization (VSS) for large-scale video analytics and the NVIDIA Cosmos™ Open World Foundation Model (WFM) for world understanding and reasoning to transform visual input into structured, actionable agent intelligence.
At the core is a digital twin-enabled training pipeline coupled with NVIDIA Physical AI Data Factory blueprints. The Physical AI Data Factory blueprint serves as a single reference architecture that unifies data curation, generation, and validation at scale. This blueprint uses Cosmos WFM, including Cosmos Transfer and Cosmos Evaluator powered by Cosmos Reason, to transform limited training data into large, diverse datasets. Linker Vision is actively testing a physical AI toolchain on Azure that integrates Blueprints with several Azure services. This approach supports scalable model development and continuous improvement of complex physical environments, generating model-ready training datasets from raw data.
To process massive real-time video streams at scale, Linker Vision is working with carriers who are building AI grids (geographically distributed and connected AI infrastructure) based on NVIDIA reference designs.
With AT&T, the system runs at edge sites that act as AI grid nodes, bringing video inference closer to cameras and sensors for lower latency and more reliable operation.
Linker Vision is working with T-Mobile to pilot distributed deployments leveraging a grid of AI-RAN-enabled infrastructure powered by NVIDIA RTX 6000 PRO Blackwell Server Edition. An ongoing smart city project in San Jose is integrating point cloud data, 3D digital twins, and vision AI for real-time transportation optimization and urban planning.
Building on this foundation, Linker Vision connects simulation, training, and deployment into a continuous AI lifecycle, bringing physical and inferential AI to real-world operations.
