All the technologies

CAFEIN provides federated evaluation tools that allow models to be tested across diverse data sources held by research institutes and experimental facilities worldwide.

CAFEIN is a comprehensive federated platform developed and hosted at CERN. Integrating specialized software components, Infrastructure-as-Code (IaC), and CERN’s robust infrastructure, it’s designed to support large-scale scientific collaboration.

CAFEIN extends beyond training with federated analytics capabilities, enabling distributed statistical analysis across sensitive datasets.

Federated learning enables privacy-preserving collaborative model training across distributed data sources without transferring any raw data.

By combining hardware and system metrics, model characteristics, regional energy data, and federation architecture details, it estimates CO₂ emissions and identifies optimisation strategies such as quantisation, sparsification, and decentralised training.

We use graph neural networks to model complex relationships within heterogeneous datasets.

Engineered for resource-constrained environments, our lightweight software models enable IoT devices to operate at their full capacity. This approach ensures optimal performance without sacrificing battery life and device portability.

Page under construction