DataMesh has announced the launch of a new embodied AI data product solution, and this marks a major step toward building smarter industrial robots. The focus keyword DataMesh Robotics appears here as it is central to the announcement. The company revealed the product on January 15, 2026, and it aims to help robot makers and robotics teams train their systems in more realistic and evolving environments. The solution allows teams to build detailed industrial scenes, simulate physical behavior, and generate large volumes of labeled synthetic data. Moreover, it solves one of the trickiest parts of embodied AI training by helping users define task goals and clear reward signals.

DataMesh Robotics stands out because it creates an “Executable Industrial Digital Twin.” Many tools only show static 3D models, but DataMesh Robotics creates scenes that can run like real industrial spaces. Objects can move, processes can change over time, and events can appear as they would on a factory floor. This dynamic setup gives robotics teams training data that matches real operating conditions more closely. As a result, teams gain a better foundation for training robots to handle complex tasks safely and correctly.

DataMesh has earned recognition from Gartner in several reports on Intelligent Simulation. This acknowledgment reflects the company’s long-term investment in digital twin and simulation technologies. With this launch, DataMesh Robotics becomes compatible with leading robotics simulation platforms, including NVIDIA Isaac Sim and Omniverse. Users can export industrial assets and training data directly to these ecosystems, and this flexibility makes it easier for companies to fit the solution into their current workflows.

The company has already completed prototype testing. It is now working with telecom operators, data labeling partners, and other enterprise teams on pilot projects. These partnerships will help expand the library of industrial assets, improve task templates, and strengthen compatibility with mainstream robotics environments.

The new solution offers several clear highlights. First, the executable digital twin allows objects, processes, events, and business logic to operate in one complete system. Second, the platform supports real industrial interaction, such as starting or stopping production lines or switching machine states. Third, teams can generate large volumes of synthetic data with automated ground‑truth labels. This includes segmentation, bounding boxes, depth maps, keypoints, and many other formats. Fourth, the platform also provides “non‑visible data,” such as temperature or pressure, which helps robots understand conditions that affect real‑world decisions.

Another key feature is its approach to reward design. In industrial environments, defining task goals and success conditions is often more difficult than building the simulator itself. Many tasks have strict safety rules, multiple steps, and limited visibility. DataMesh Robotics solves this by offering a low‑code method for setting objectives, constraints, and reward structures. Teams can use industry‑specific metrics like force thresholds or inspection completion rules to create stable and repeatable training setups. With the digital twin market expanding rapidly as industries adopt advanced simulation and automation, platforms like this are increasingly critical for efficient and safe industrial AI deployment.

The system integrates with major simulation platforms and supports on‑premises, private cloud, and hybrid cloud deployments. This flexibility helps organizations maintain control over data and security while still advancing their AI capabilities.

DataMesh Robotics aims to support robot manufacturers and robotics application teams. Specifically, these teams need to train robots for assembly tasks, warehouse navigation, facility inspections, as well as operations in dangerous environments. In addition, the platform helps them cover long‑tail scenarios, thereby improving safety and, moreover, speeding up deployment cycles.

DataMesh CEO Jie Li said the goal is to build a training world that changes like the real world. He explained that the company offers not only detailed scenes and synthetic data, but also a dynamic simulation layer that evolves with workflows and events. This allows teams to run the full training cycle more effectively.