The Federated  Learning Platform

Leverage decentralized business sensitive and regulated data in your machine learning pipelines.

Our vendor agnostic software platform lets you take machine learning to data instead of data to machine learning. Develop regulatory compliant solutions for computer vision, predictive maintenance, anomaly detection and more.

Book a demoShow me the code

Clients & Collaborations

Unlock value from siloed data with federated learning

The most valuable data in the Enterprise can often not be pooled in a central cloud for regulatory or practical reasons.

By leveraging data distributed in multiple data silos or over large numbers of devices, better predictions can be achieved.

Scaleout's software platform gives your data scientists the central cloud's unified development experience without the need to centralize data. This greatly reduces security risk in industrial machine learning.  

Learn more
  • Remove regulatory barriers in ML
    Enable collaboration between organisations and subsidiaries by removing regulatory barriers to data access (GDPR, sensitive data).

  • Reduce latency and cost in EdgeAI Train models and make predictions directly on edge nodes and devices. Avoid costly data transfers to a central cloud (IoT, connected vehicles).

  • Improve data privacy and security
    Scaleout platform reduces risks in AI projects by removing the need for copying and moving datasets for model training.

Scaleout Studio = FedML + MLOps
Scaleout's approach is both vendor agnostic and machine leaning framework agnostic. Studio is a SaaS that can be deployed on any public or private cloud (AWS, Azure, OpenStack, Kubernetes cluster).
Ephicient logoOE logo2020INC logoThe Paak logoAriseHealth logo

How does Federated Learning work? 

  • A data scientist provides a machine learning model definition. This "compute package" defines how to read in local data and how to compute local model updates.
  • The federated learning runtime connects client sites / devices with one or many aggregation servers. The compute package is distributed to each client. Client authentication, authorization and code distribution is handled in production by Scaleout Studio.
  • Training updates are computed locally on each client. Data never leaves its original site - only model parameters are exchanged for aggregation. This process is iterated until convergence to a global, federated model.

Learn more:

FEDn: open sourcefederated learning

Get started developing FL solutions today using our open source framework FEDn. Develop solutions for computer vision, NLP, fraud detection and more, with world-leading scalability and resilience.

Apache2 with community support in Discord.
Use Tensorflow, Torch, sklearn, or any machine learning framework.
A clear upgrade path to production and SaaS with Scaleout Studio.

Join the conversation!

The Scaleout community is focused on decentralized AI and federated learning.

Our focus is on solving decentralized AI problems using the open source project FEDn, but other related projects and subjects are also discussed.

Latest updates