Meet Sereact: The AI Startup Reinventing Warehouse Robots

What is Sereact?

Sereact is a pioneering German AI and robotics developer that has fundamentally transformed the landscape of industrial automation since its founding in 2021. Headquartered in Stuttgart-Vaihingen, the company focuses on creating highly adaptable, hardware-independent artificial intelligence models tailored for autonomous pick-and-place robots. Historically, the robotics industry has been severely constrained by traditional training methods, such as reinforcement learning (RL), which is highly resource-intensive and relies on endless trial-and-error, or imitation learning (IL), which demands extensive human-led instruction.

Sereact bypasses these outdated methodologies by introducing a software-first approach focused on embodied AI. The core mission of Sereact is to provide a “robotic brain” that can seamlessly retrofit into the world’s vast fleet of existing industrial robots, transforming them from rigid machines into highly intelligent systems capable of adapting to novel environments without any prior programming. By prioritizing flexible automation, the company allows warehouses, logistics centers, and manufacturing plants to rapidly scale their intralogistics processes and tackle the challenges of ever-changing operational demands.

Who Founded Sereact?

The visionary minds behind Sereact are Co-Founder and CEO Ralf Gulde and Co-Founder and CTO Marc Tuscher. The two innovators have a long-shared history, having known each other since their school days when they collaborated on early engineering projects. Gulde’s passion for robotics was ignited at age ten with a Lego Mindstorms kit, eventually leading him to study mechatronics, while Tuscher pursued computer science.

The foundational idea for Sereact took root while the duo was completing their master’s degrees and subsequent doctorates at the University of Stuttgart’s Institute for Control Engineering of Machine Tools and Manufacturing Units (ISW). Recognizing a glaring lack of adaptable AI solutions in the industrial robotics market, they decided to build their own. Demonstrating an incredibly resourceful and entrepreneurial spirit, the founders literally drove around Stuttgart buying used gaming graphics cards off eBay Classifieds to construct their very first server racks. Supported by the university’s technology transfer initiative (TTI GmbH) and an EXIST start-up grant, they officially launched Sereact in 2021.

How Sereact PickGPT Works

At the heart of the company’s software revolution is Sereact PickGPT, a groundbreaking Robotics Transformer that represents the world’s easiest tool for instructing and programming robots. PickGPT brilliantly merges the advanced reasoning capabilities of large language models (LLMs)—similar to the technology powering ChatGPT-with the company’s proprietary, patented computer vision models.

The technical architecture of Sereact PickGPT is highly sophisticated. It begins by fusing multimodal sensor data, such as RGB and depth inputs, and dividing this information into fixed-size patches. These patches are embedded linearly and positionally before being fed into a Vision Transformer (ViT), which handles critical tasks like segmentation and scene decomposition. The LLM simultaneously tokenizes textual input, and a cross-attention mechanism bridges these text concepts with specific image sections. This process, known as “visual grounding,” enables PickGPT to accurately recognize and grasp an “open-set” of previously unseen objects simply through knowledge transfer.

By processing natural language, Sereact PickGPT enables warehouse employees to command robots using simple voice instructions or intuitive user interfaces, entirely eliminating the need for complex coding skills. Furthermore, the system is equipped with built-in failsafe strategies; if a robot fails to grasp an item on the first try, the LLM reassesses the physical environment, dynamically adapts its strategy, and tries again, ensuring continuous operational efficiency.

The Sereact Cortex 2.0 Brain

While PickGPT introduced natural language to robotics, the Sereact Cortex 2.0 Brain represents the next generation of physical AI capabilities. The foundational difference between early autonomous systems and Cortex 2.0 is the shift from a “try-and-see” approach to a highly calculated “plan-and-try” methodology. Sereact Cortex 2.0 augments a vision-language-action (VLA) model with a predictive “world model.” Before a robotic arm even moves, Cortex 2.0 generates multiple candidate trajectories, tests them against a learned model of physics, and scores each outcome for risk, stability, and efficiency.

Because Sereact Cortex 2.0 plans in a “visual latent space” rather than relying on hardware-specific joint commands, the exact same neural brain can control single-arm picking cells, dual-arm return stations, fixed cells, and even humanoid robots. The most significant advantage of Cortex 2.0 is its compounding data flywheel. The model has been trained on over 1 billion real production picks gathered from real-world warehouse deployments. Every successful pick, failure, and recovery across the entire global fleet is captured, filtered, and used to continuously retrain the centralized model. As a result of this massive, real-world data infrastructure, a Sereact system currently requires remote human intervention only once in roughly every 53,000 picks.

Warehouse Automation & Logistics Use Cases

The technology deployed by Sereact is designed to solve some of the most complex bottlenecks in dynamic supply chains. One of the most prominent use cases for Sereact is intelligent, zero-shot item picking. Because the AI evaluates shape, material, and color in real time without manual data acquisition, it can instantly manipulate delicate materials, transparent packaging, and constantly changing SKUs.

Another revolutionary use case is simplified returns processing, which is notoriously labor-intensive for e-commerce retailers. Using Sereact PickGPT, the robot can independently differentiate between discarded packaging materials and the actual returned products. By analyzing textual descriptions and visual product images, the system can automatically sort items without requiring a barcode scan, drastically accelerating the entire reverse logistics pipeline. Additionally, Sereact technology acts as an automated quality control system; it cross-references real-time visual feeds against product databases to immediately identify defective products, incorrect items, or blacklisted materials.

Sereact Funding & Investors

To fuel its rapid technological advancements and global expansion, Sereact has successfully attracted top-tier venture capital, securing over $140 million in total funding. In January 2025, Sereact closed a €25 million Series A funding round led by Creandum, with participation from Point Nine, Air Street Capital, and notable angel investors, including former Formula 1 World Champion Nico Rosberg.

The company’s momentum accelerated further in April 2026 with a massive $110 million Series B round led by Headline. Other major institutional investors, such as Bullhound Capital, Daphni, and Felix Capital, also participated in this round. The immense influx of capital is earmarked for two primary strategic goals: scaling the computing power necessary for the Sereact Cortex 2.0 world model and aggressively expanding operations into the United States. To support this North American market entry, Sereact has opened its first US office in Boston, actively hiring commercial, application, and engineering experts. The funding also supports the company’s foray into humanoid robotics and more complex manufacturing tasks.

Sereact Partners and Customers

Today, Sereact is recognized as the most deployed AI picking robot company globally, boasting over 200 live systems actively operating across Europe. The company achieves this scale through strategic integration partnerships, such as their collaboration with AWL, a global automation provider with over 4,000 integrated robots. By combining AWL’s mechanical engineering expertise with the Sereact AI model, the partnership is executing a worldwide rollout of flexible intralogistics systems.

The tangible business impact of Sereact is best illustrated by its logistics customers. At the MS Direct fulfillment site in Arbon, Switzerland, a Sereact robot navigates a highly heterogeneous AutoStore system containing over 60,000 different SKUs. The robot flawlessly processes around 1,500 single-pick orders daily and runs during night shifts to handle preparatory tasks, resulting in a full return on investment (ROI) in just nine months. Similarly, the fulfillment expert Active Ants implemented a Sereact AI-driven robotic arm in Dorsten, Germany, which successfully picks 600 parcels per hour with a staggering 99.99% accuracy rate, allowing full orders to be packed in 15 minutes without human intervention. Due to this success, Active Ants has already installed a second Sereact unit in Roosendaal, the Netherlands. Beyond logistics, the company’s embodied AI is also trusted by major global automotive and retail brands, including the BMW Group, Mercedes-Benz, Daimler Truck, Bolt, and PepsiCo.

---