News
Warsaw's Nomagic Unveils AI That Halves Warehouse Robot Help Calls
Warsaw-based robotics company Nomagic has deployed a vision-language-action model at paying customer sites that meaningfully cuts how often warehouse robots need to call for human help. It's one of the world's first commercial deployments of such a model outside the lab.
Nomagic, a robotics company with its European headquarters in Warsaw, announced that its new artificial intelligence model for controlling warehouse robots is now running at paying customers and has cut in half the number of situations in which a machine has to call for human help. It's one of the world's first deployments of a vision-language-action model in real, commercial production, rather than in a lab or a demo showcase.
What Was Announced
Nomagic built a research lab focused on a single task: teaching a warehouse robot to handle so-called edge cases, the unusual situations in which the machine previously got stuck and had to call an operator. The vision-language-action model perceives its surroundings, understands text instructions, and takes physical action based on them, for instance deciding how to grip an unusually placed product.
The key difference from most competitors' approach lies in the order of work. Instead of building a general, universal robot brain and only then looking for applications for it, Wulfmeier's team started with a specific, narrow problem in working warehouses and developed the model around it.
Most of our community is racing to build the most general robot brain possible. We're betting that the harder part is real mastery of the task - Markus Wulfmeier, Chief Scientist at Nomagic
Why It Matters
In warehouse robotics, reliability matters more than flashy demos. Logistics companies only accept automation when errors occur less often than with human labor, and every robot downtime means cost and delay in the supply chain. Until now, vision-language-action models have mostly been tested in lab conditions or in short, controlled pilots.
Nomagic says its VLA doesn't yet operate independently at the 99.9 percent reliability level the industry demands, so the model has been wrapped in classic, deterministic robotics software that acts as a safeguard. This layer catches the model's errors before they affect warehouse operations, which already lets the company offer customers guaranteed reliability for the system as a whole.
The bar in the physical world is high: 99.9 percent isn't a marketing trick, it's the cost of admission to the building - Kacper Nowicki, co-founder and CEO of Nomagic
Polish Base, Global Customers
Nomagic has its European headquarters in Warsaw and its US headquarters in Sandy Springs, Georgia, but its customers operate across Europe. They include Brack.Alltron in Switzerland, where the company first fully deployed the new model, and Zalando, for which its robots perform millions of product picks per month.
Roland Brack, founder of Brack.Alltron, described the effect of the deployment as a breakthrough in how machines understand their surroundings, letting the company run autonomous night and weekend shifts without staff present.
Today we're seeing robots that truly understand their environment. That intelligence lets us run autonomous shifts through nights and Sundays - Roland Brack, founder of Brack.Alltron
What's Next
Nomagic hasn't disclosed the funding behind the new lab or the pace at which it plans to expand deployments to more customers. The company says its next step will be gradually moving from mastery of single, narrow tasks toward a more general system covering an increasingly broad range of warehouse operations.
For Poland's tech sector, it's a signal that domestic teams can attract researchers from the world's leading labs and compete in one of the most demanding fields of artificial intelligence, physical robotics operating in real business conditions rather than just in simulations.
Sources: Fortune (fortune.com), TheNextWeb (thenextweb.com)
