NATO-backed defense startup Scaleout Systems is deploying small machine learning models on military drone hardware to perform target detection, identification, and selection on edge devices. Originally founded in 2018 by Uppsala University researchers for commercial vehicles, the Swedish company pivoted to defense application following the 2022 invasion of Ukraine. Scaleout was subsequently selected for NATO’s Defence Innovation Accelerator for the North Atlantic (DIANA) Challenge Program in 2025.
According to Scaleout CEO Andreas Hellander, the company bypasses centralized frontier models from labs like OpenAI in favor of compact computer vision models running directly on drone processors, pilot tablets, and field command nodes. The system utilizes federated learning techniques, allowing edge devices to gather sensor data, share intermittent updates with local platoon servers, and continuously retrain models to adapt to shifting tactical environments like urban combat.
This decentralized approach addresses critical vulnerabilities created by electronic warfare signal jamming and attacks on central data centers. By enabling autonomous operation without continuous cloud communications, Scaleout offers military operators sustainable battlefield intelligence and reduced dependency on centralized infrastructure.
Why it matters
Edge-deployed AI models enable real-time computer vision without relying on cloud connectivity or high-bandwidth links.
Federated learning allows local model retraining on field data, solving domain adaptation issues across changing physical environments.
Defense tech startups are capitalizing on tactical requirements for offline resilience against electronic warfare and jamming.
Source: arstechnica.com
