Sereact is building vision-language-action models into warehouse robot picking
A January 20, 2025 Series A, named warehouse customers, and PICKGPT give Sereact a software-led robot picking anchor.
Sereact said on January 20, 2025 that it raised ?25 million in Series A funding. Creandum led the round alongside Point Nine and Air Street Capital, and the company named Daimler Truck, Bol, MS Direct, and Active Ants as customers.
Those customer names give Sereact more than a foundation-model pitch. Warehouses need robots to handle changing SKUs, packaging, bins, and instructions without writing fixed code for every item. Sereact?s Vision Language Action Models are aimed at that gap between perception, language, and robot behavior.
Sereact describes VLAMs as hardware-agnostic models for robots in warehousing and logistics. PICKGPT adds a natural-language interface for robot tasks, so operators can instruct the system without building a new automation sequence for every exception.
Sereact was founded in Stuttgart by Ralf Gulde and Marc Tuscher. The company sits in the overlap between warehouse picking software, robot perception, and physical AI, where model quality has to show up as successful picks rather than benchmark language alone.
The competitive field includes Covariant, Physical Intelligence, Skild AI, RightHand Robotics, Ambi Robotics, Osaro, Mujin, and internal robot-learning teams at large warehouse automation companies. Sereact?s distinction is applying vision-language-action models directly to warehouse manipulation with named commercial customers.
Public material does not show paid deployment count, robot count by customer, pick success rate by SKU class, autonomy intervention frequency, commissioning time by integrator, pricing, customer retention, renewal rate, or repeat deployment rate. The proof is funding, named customers, and product direction.
Sereact?s strategic test is whether physical AI can improve warehouse picking before humanoids or general robots arrive at scale. If VLAMs and PICKGPT reduce reprogramming and exception handling in real cells, Sereact can become a software layer for robot picking across hardware platforms.
Have a robotics update Korthos should review? Send news, deployments, product releases, funding rounds, research, or media to tips@korthos.xyz or reach out on X at @agkorthos.