Uppsala innovation in cyber security to help NATO

“It feels important that we can contribute, given how the world has developed,” says Andreas Hellander, CEO of Scaleout and researcher at the Department of Information Technology.

One of the biggest challenges in a defence context is being able to communicate without sensitive information falling into the wrong hands. Today, distributed sensors generate a wide range of surveillance data, which can be used to train AI systems. However, in current workflows this data must be sent from the edge devices to a central server, explains Hellander.

“This introduces both security risks and practical blockers in contested network environments. Scaleout has developed a solution that allows local actors to maintain strict control of their local sensitive data, yet contribute to a shared machine learning model.”

Proactive security measures

The technology is called federated machine learning and it allows multiple users to jointly train models without centralising raw data. Looking beyond defence applications, electronic tools are an inevitable part of our everyday lives. Phones, laptops and smart appliances generate large amounts of data. While this data holds great potential for improving machine learning models, it often contains sensitive information, researcher Salman Toor at the Division of Scientific Computing points out.

“Protecting privacy and security is essential. Security and privacy measures can no longer be reactive; they must be proactive – preventing damage before it occurs.”

For about ten years, Hellander and Toor have been researching systems that combine large-scale data processing in the cloud with compute and data storage closer to the data sources. In 2017, a business idea began to take shape and, with the help of UU Invest, they were able to finance a software prototype. Together with co-founders with backgrounds in startups and business, Hellander started Scaleout together with Toor as Chief Technology Officer.

Great need in mission-critical organisations

Since then, the company has grown to 15 employees. Today, Scaleout develops edge AI and federated machine learning technology with a particular focus on defence, the automotive sector and smart manufacturing. There is a great need to leverage distributed training data in mission-critical organisations – this is also crucial at the tactical edge of the battlefield where conditions change rapidly, says Hellander:

"New threats, changing landscapes and shifting conditions require models that can be adapted in real time. Without updates, critical systems that rely on machine learning risk losing their relevance when they are needed most.”

This is one of the reasons why Scaleout was selected for NATO's innovation accelerator in February 2025. Together with five other Swedish companies, the company is part of NATO's DIANA Challenge Programme with the FEDAIR (Federated Aerial Intelligence for Recon) project. The programme brings together products and technologies that can be used both for commercial and military purposes, known as ‘dual use’.

“It’s a game-changing approach”

According to Toor, modern defence solutions must be fast, agile, and built on resilient edge units.

“This architecture gives operational teams a tactical advantage, but also introduces the complex challenge of coordinating multiple edge nodes in fast-changing environments,” he says and continues:

“Smart, decentralised coordination is key to achieving reliable outcomes. The Scaleout Edge platform delivers exactly that – intelligence at the edge with confidence, without relying on massive raw data transfers. It’s a game-changing approach that scales seamlessly and unlocks powerful applications across defence and beyond.”

One of Scaleout’s strengths is the company's ability to benefit from both basic research and industrial insights, adds Hellander.

“The exchange of knowledge helps both the company and the research group at Uppsala University to develop.”

Anneli Björkman