Reply at NVIDIA GTC Berlin 2026: Physical AI learns from ongoing operations

Reply [EXM, STAR: REY] will be present at NVIDIA GTC from 20 to 22 October 2026 at STATION Berlin, showcasing how physical AI – artificial intelligence for robots and other systems that interact with their physical environment – can be continuously improved using operational data. At stand #7023, Reply’s experts will demonstrate how virtual development and industrial deployment work together. 

Digital representations of machines, equipment and production environments play a central role in this approach. Using NVIDIA Omniverse, Reply brings together data from different systems in digital twins and replicates real-world operations in virtual environments. This allows new configurations, processes and robotics scenarios to be tested. 

The intelligent robot coordination project developed with the Otto Group and NVIDIA demonstrates how this works in practice. For this project, Roboverse Reply created a highly precise digital twin of a warehouse environment. It replicates different robotic systems, their movements and their interactions in a virtual environment. New robots, layouts and processes can be tested within the digital twin before being introduced into the physical warehouse. It also provides the basis for centralised coordination of different robot fleets. 

In addition to robotics and digital twins, Reply uses NVIDIA GPU acceleration to develop and scale AI applications. The Reply Model Factory offers a standardised method for creating specialised generative AI models by combining proprietary company data with domain expertise. The Model Factory covers the entire model lifecycle - from data preparation and training to validation, deployment, and continuous optimisation - enabling organisations to create models tailored to their specific business requirements.  

Further live demos will focus on collaboration between people and machines. Visitors can see how digital humans tailor their responses to individual questions and context, providing information in multiple languages. They can serve as brand ambassadors in shops or as digital information assistants at events and exhibitions. One example is Futura, the virtual tour expert developed by Reply for use on board Costa Crociere’s Costa Smeralda. It helps passengers plan excursions in six languages, recommends tours based on their preferences and allows them to book directly. 

On 21 October at 2:00 p.m., AWS, NVIDIA and Reply will also discuss how to accelerate the transition from proof of concept to production in a fireside chat. Mike Groocock (AWS), David Boyd (NVIDIA) and Maxim Stoljar (Roboverse Reply) will share practical insights from their joint work. 

Kai Uwe Ernst, Executive Partner at Reply, commented: “With physical AI, it is no longer enough to implement a single use case at a technical level. What matters is bringing models and applications reliably into production, refining them during operation and applying them to new scenarios. To achieve this, we combine our expertise with NVIDIA technologies.” 

What drives all this is the partnership between Reply and NVIDIA. For more information: www.reply.com/en/nvidia