assumptions hold. Data cannot move because of privacy law, classification boundaries, or bandwidth costs. Networks are contested, intermittent, or fully denied. And a model trained on averaged central data washes out the local patterns that make each node's environment distinct.

Our platform, Scaleout Edge, is built for these conditions. It manages the full lifecycle of AI models across decentralised systems (defence platforms, vehicle fleets, industrial sites, and regulated data silos) covering training, deployment, inference, and governance from a single control plane.

Federated learning is the core mechanism that allows models to keep improving without moving sensitive data: models go to the data, train locally, and only weight updates return. Equally central is what happens around training, secure over-the-air model distribution, on-device inference that continues through connectivity loss, per-node model personalisation, fleet-wide observability, and an immutable audit trail across every node. The result is AI that stays operational, accurate, and governable in the field, with sovereignty enforced architecturally rather than by policy.

On top of the core platform, capability modules deliver domain-specific tooling for computer vision, drone and autonomous systems, speech recognition, and adversarial ML security.

Scaleout Edge has been demonstrated in defence, automotive, space, and industrial settings. Customers and partners include BAE Systems Bofors, Saab, Scania, FMV, BMW, Oracle, and Akkodis. The company was selected as a NATO DIANA Innovator in 2025.

Scaleout was founded in 2018 as a spin-out from Uppsala University, originating from research in large-scale distributed systems and scientific computing. The team of 20 includes seven PhDs and two associate professors, combining research depth with industry experience from Google, Saab, BCG, Oracle, and MySQL.