Italian technology consultancy Reply has earned NVIDIA’s Physical AI Specialization, a designation recognising its capabilities in developing and deploying AI-driven robotics and digital twin solutions. The recognition highlights a growing focus on connecting simulated environments with industrial operations, where companies want to test robotic systems virtually before introducing them into warehouses, factories and logistics networks.
Reply’s work with Otto Group at a German fulfilment centre illustrates the approach: a virtual representation of warehouse operations provides a foundation for coordinating autonomous mobile robots from different manufacturers alongside existing warehouse systems.
Digital Twins Connect Simulation With Warehouse Operations
The specialization follows a technical assessment that covered a customer implementation, solution architecture, certified OpenUSD expertise and Reply’s development roadmap. It places the company among the first NVIDIA partners to receive the designation, according to Reply’s announcement published October 19, 2026.
The underlying challenge is coordinating robots that may come from different manufacturers and operate through separate fleet-management systems. Without a common coordination layer, integrating new machines or changing warehouse workflows can require substantial engineering and testing in live facilities.
Reply’s work with Otto Group addresses that problem through a high-fidelity digital twin of the Hermes Fulfilment logistics centre in Loehne, Germany. The virtual representation provides the basis for Otto Group’s Robotic Coordination Layer, developed by technology and consulting provider Otto Group one.o.
The system connects autonomous mobile robots from different manufacturers with fleet-management software and the warehouse management system. Using NVIDIA Omniverse and NVIDIA Isaac, the teams can simulate new layouts, robotic configurations and workflows before physical deployment.
The digital twin was built using spatial data captured with scanner and camera equipment mounted on Boston Dynamics’ Spot robot. That data was processed into three-dimensional models intended to reflect the warehouse environment and its operational constraints. The virtual environment can then support simulation, process analysis and the evaluation of potential changes.
The project is being piloted at the Loehne facility and is intended to provide a blueprint for expansion to additional Otto Group locations. The announcement does not disclose independent performance measurements or quantify the project’s realised cost savings.
Why Physical AI Requires More Than a Model
Physical AI describes AI systems that interact with the real world through machines such as robots. Unlike software agents operating entirely in digital environments, these systems must account for spatial constraints, sensor inputs, movement and the consequences of physical actions.
That makes simulation and system integration important parts of deployment. A model that performs well in a controlled demonstration may still encounter unexpected conditions in a busy warehouse. Digital twins let engineering teams evaluate some scenarios virtually, although the quality of those simulations depends on how accurately they represent the physical environment.
Reply combines robotics integration, reality capture, digital twins and sensor technologies with NVIDIA’s simulation and AI infrastructure. The specialization recognises the company’s capabilities in bringing those components together, rather than announcing a new foundation model or standalone robotics product.
For manufacturers and logistics operators, the commercial question is whether such systems can simplify deployment across different robot fleets while maintaining safety, reliability and compatibility with existing operational software.
Enterprise Deployment Remains the Test
Reply CTO Filippo Rizzante said Physical AI requires simulation, robotics, sensor data and software engineering to work together in architectures designed for real operating environments. The company positions its NVIDIA partnership around that integration work, from virtual development and AI training to production rollout.
The approach could help organisations test changes before applying them to live operations and reduce some of the disruption associated with physical experimentation. However, simulation cannot eliminate the need for on-site validation, safety procedures and monitoring after deployment.
Reply’s recognition comes as industrial AI initiatives increasingly connect virtual models with real equipment and operational data. Its Otto Group project provides a concrete example of how digital twins can serve as a coordination layer for multi-vendor robotics.
The next measure of success will be how reliably the system performs under real warehouse conditions, how easily it accommodates additional robot types and whether the approach can be replicated across sites without extensive custom engineering.
Market Landscape
Industrial AI is moving beyond isolated automation projects toward systems that combine robotics, simulation and operational software. NVIDIA Omniverse provides technologies for building and simulating virtual environments, while NVIDIA Isaac supports robotics development and simulation.
Reply’s specialization highlights the role of systems integrators in connecting these technologies to existing enterprise operations. The distinction matters because industrial deployments depend not only on AI capabilities, but also on data quality, hardware integration, safety validation and operational change management.
Top Insights
- NVIDIA recognition: Reply received the Physical AI Specialization following a technical assessment of its expertise and implementation capabilities.
- Warehouse use case: The Otto Group project uses a digital twin to support coordination of robots from multiple manufacturers.
- Virtual validation: NVIDIA Omniverse and Isaac support simulation of warehouse layouts and robotic workflows before deployment.
- Integration focus: Reply combines robotics, spatial data capture and enterprise system integration rather than introducing a new foundation model.
- Deployment evidence matters: Independent performance metrics and results from broader rollouts will be needed to assess the solution’s operational impact.
Power Tomorrow’s Intelligence — Build It with TechEdgeAI
