Teradyne Robotics and Sereact partner to automate retail returns

Real-world deployment at Zalando

The technology is no longer confined to laboratory settings or small-scale pilots. Sereact and Universal Robots are already deploying the system with CEVA Logistics to support Zalando, a major European fashion and lifestyle platform. The robots are currently processing returned fashion items at fulfillment centers in Greven, Germany, and Świebodzin, Poland.

These specific locations serve as the first phase of a larger plan to scale returns automation across the continent. Because returned items rarely arrive in a predictable sequence or packaging, the robots must be able to adapt to different shapes and orientations. The system is designed to recondition these goods quickly so they can be sold again.

This automation also helps retailers meet specific legal requirements. The European Union has regulations that prohibit the destruction or landfill disposal of returned merchandise. By using AI to sort and prep these items, companies can keep products in the supply chain while staying compliant with environmental laws.

The 'one brain' architecture

A key technical advantage of the Cortex platform is its unified production model. Dr. Ralf Gulde, co-founder and CEO of Sereact, explained that one brain serves every robot without requiring retraining for every new item or task. This means that once a robot learns how to handle a specific type of object, every other robot in the fleet benefits from that knowledge immediately.

The system is designed to handle a wide variety of industrial use cases beyond just returns. It can be applied to piece picking, sortation, order induction, machine tending, bin consolidation, kitting, and packing. Because the software handles mixed goods and unknown items, it can be integrated into many different types of manufacturing and logistics environments.

To support this global rollout, Sereact recently opened a U.S. headquarters in Charlestown, Massachusetts. This move is intended to meet the growing demand for these AI-powered solutions in North America. The partnership leverages Teradyne's existing global distribution and integrator network to bring this technology to a wide range of customers worldwide.

Training and hardware specs

The development of the Cortex platform relies heavily on NVIDIA's technology stack. The AI is trained in NVIDIA Isaac Sim, which is an open simulation framework. It also utilizes NVIDIA Cosmos open world models to prepare the robots for the complexities of real-world environments. Before any model is actually deployed, it is post-trained using reinforcement learning via the NVIDIA Isaac Lab framework.

The hardware setup is also specifically optimized for speed and validation. In active deployments, Cortex runs on NVIDIA Jetson Thor at the specific cell where inference must happen at line speed. For site-level development, validation, and model updates, the system uses NVIDIA DGX Spark. This tiered approach allows for high-speed operation on the factory floor while maintaining a robust backend for updates.

Teradyne Robotics is also contributing to the broader AI ecosystem by sharing assets. They are contributing SimReady OpenUSD robot assets to the SimReady Central Library on Hugging Face. These assets include models for both Universal Robots and Mobile Industrial Robots, which helps other developers discover and reuse robot models in simulations with less preparation time.

Scaling to new markets

The partnership was officially announced during the NVIDIA GTC Berlin 2026 event. Both companies emphasized that their goal is to move beyond the demo stage and into high-volume, useful deployment. Jacob Pascual Pape, the chief commercial officer of Teradyne Robotics, noted that their robots are built for the AI era because they are open, connected, and proven through over 100,000 industrial deployments.

By combining Sereact's software with Teradyne's hardware, the companies aim to scale fast across various industries. The goal is to provide a reliable way for retailers and logistics providers to adopt AI-powered automation. This includes handling the most difficult jobs in the supply chain, such as managing the unpredictable flow of returned goods.

As the partnership matures, the focus will shift toward expanding the reach of the 'one brain' model. By making it easier for companies to deploy robots that can handle unknown items, the companies hope to standardize how warehouses manage inventory. This could lead to more flexible logistics hubs that can switch between different types of products with minimal downtime.