Machine learning that trains where the data lives. On drones, vehicles, and industrial fleets. Raw data never leaves the device.
Winter Demo 2026 | BAE Systems Bofors
Scaleout enabled autonomous drones to learn and operate in contested Arctic environments without central data links.
Real-time AI model distinguishes multiple target types locally during autonomous search flights.
Onboard prioritization logic selects targets based on mission requirements and real-time environmental data.
Full autonomy without reliance on external processing or continuous radio links for flight path adjustment.
Scaleout Edge Platform
The AI systems that matter most run where data cannot be centralised and networks cannot be trusted. Scaleout Edge is designed for those environments.
Models train on-device. Only encrypted weight updates are shared. Your data stays exactly where it is, enforced by architecture.
Every version recorded with its compute package hash and session lineage, creating a complete audit trail across the fleet.
Training metrics and hardware telemetry from every node. Visibility where you have no physical access.
Extends your existing ML stack to handle distributed model training and management across devices.
Edge AI Modules
Ready-to-use modules for common edge AI workloads built on the Scaleout Edge platform. Each includes reference models, training workflows, and deployment templates so teams can move from prototype to deployment quickly.
Perception Module
Federated-ready vision models for detection, classification, and segmentation that improve continuously through local fine-tuning without sharing raw data.
Drone & Autonomy
Base models and integration tools for autonomous platforms with PX4 and TAK support, featuring on-device inference and telemetry caching for offline use.
Adversarial Modeling Toolkit (AMT)
Test federated models for privacy leaks with tools for inversion and membership attacks, plus simulations for data poisoning, backdoors, and gradient inversion.
ASR (Speech & Language)
Federated fine-tuning of speech and language models on private data, with Whisper and transformer workflows optimized for edge devices like NVIDIA Jetson.
# Configure ASR for secure edge deployment
import scaleout_edge as soe
from transformers import WhisperForConditionalGeneration
# Load base model for fine-tuning
model = WhisperForConditionalGeneration.from_pretrained("openai/whisper-large-v3")
# Initialize Federated Fine-tuning
trainer = soe.FederatedTrainer(
model=model,
dataset="/data/local_voice_records",
strategy="fed_avg"
)
# Compile for hardware target
optimized_model = soe.compile(
target="nvidia_jetson_orin",
precision="int8",
enable_tensor_rt=True
)
optimized_model.deploy(port=8080)
Use Cases
From contested defense environments to regulated industries, a consistent pattern: distributed data that cannot move and AI that must keep improving.
Drones, unmanned systems, and forward-deployed sensors operating in contested or bandwidth-denied environments.
Thousands of vehicles generating terabytes of telemetry daily — too much to centralize, too valuable to ignore.
Remote mines, railways, and energy infrastructure where data sovereignty is a legal requirement.
High-value datasets in government and healthcare siloed by privacy law and classification.
Technical publications and perspectives from the Scaleout team.
Autonomy is not a single AI model; it is a tightly coupled system of hardware-agnostic modules. Our stack transforms raw visual input into structured mission logic, enabling drones to maintain persistent target memory and execute complex maneuvers in communication-degraded environments.
Sigvard Dackevall
Machine Learning Engineer
We tested privacy risks against production-grade vision models like YOLO and ViT. In realistic settings, meaningful image reconstruction often collapses.
Viktor Valadi, Mattias Åkesson, et al.
Test your models against 10,000 drifting, dropping, and reconnecting clients. Scaleout Edge provides the modular infrastructure to move from lab prototypes to high-concurrency production deployments.
Scaleout