[ELASTIC Demo Series] Federated Learning Toolbox – Privacy-Preserving Collaborative Analytics

The Federated Learning Toolbox, developed by Zentrix Lab within the ELASTIC Project, enables portable cloud-edge federated learning orchestration using WebAssembly workloads across heterogeneous industrial edge environments. Industrial edge environments consist of devices with different hardware platforms, operating systems, and runtime constraints. The toolbox uses WebAssembly as a lightweight and portable execution model, allowing workloads to run consistently across distributed cloud-edge infrastructure without exposing sensitive local data. The demonstrator focuses on collaborative learning across connected factory environments. Each factory performs local training on its own operational dataset and sends only model updates to the federated learning coordinator. In this use case, the global model is trained for industrial energy consumption prediction based on operational conditions such as load type, time, and energy-related signals. The architecture separates the system into modular components: a coordinator managing orchestration, aggregation, and experiment state; shared type definitions for interoperable schemas; and edge-side WebAssembly clients executing portable, isolated workloads across heterogeneous devices. Federated aggregation uses FedAvg across multiple rounds, with real-time monitoring of convergence metrics and client participation. The demo showcases a live federated learning experiment dashboard and an interactive smart factory energy prediction interface, where operational parameters can be adjusted in real time to visualise predicted daily energy profiles. Developed by: Zentrix Labs (ZEN) Learn more: https://elastic-project.eu Funding: ELASTIC project has received funding from the Smart Networks and Services Joint Undertaking (SNS JU) under the European Union's Horizon Europe research and innovation programme under Grant Agreement No. 101139067.