Federated Learning — Docs

In regulated industries like pharma, or security-sensitive operations like offshore oil, centralizing data isn't just risky — it's often not allowed. Federated learning lets you benefit from multi-site intelligence while staying fully GDPR-compliant.

How It Works

Local training — each site trains models on its own data
Gradient sharing — only model weight updates (not raw data) are exchanged
Aggregation — a central coordinator merges the updates into a global model
Distribution — the improved global model is sent back to each site

Benefits

Privacy — raw operational data never leaves the site
Scale — models improve from collective experience across all sites
Compliance — meets GDPR, data sovereignty, and industry-specific regulations
Bandwidth — only kilobytes of model updates transmitted, not gigabytes of sensor data

Configuration

federated:
  enabled: true
  coordinator: https://fed.supi.ai
  site_id: SITE-NORTH-SEA-01
  training:
    local_epochs: 5
    aggregation_rounds: 10
    min_sites: 3
  privacy:
    differential_privacy: true
    noise_multiplier: 0.1
    clipping_norm: 1.0