WINNIIO presents the smart-building project. Federated heating control and self-learning AI across VÖFAB's buildings on the EdgeLake fabric, with full data sovereignty.
Daniel Radza of Winniio (about three years with the company) walked through the smart-building deployment Winniio has been running on EdgeLake, focused on the self-learning smart-heating digital-twin that controls building temperature with an edge-native strategy.
The previous architecture had three problems: scaling across customers (each new building tended to need its own cloud setup), data backup on pure-edge deployments, and bidirectional control of the actuators. Heating is not a read-only problem. The system needs both AI-controlled output and reliable manual override from the mobile app, which means data has to flow back from the interface to the actuator just as cleanly as it flows from the sensor to the dashboard.
EdgeLake replaced the per-customer cloud silos with a federated layer. Data stays local in each building, the operator nodes register themselves on the shared metadata layer, and the cloud (when the customer opts in) acts as an aggregator over the federation rather than as a separate data store per tenant. Backup, security, and bidirectional command flow all sit on the same fabric.
The intellectual core of Radza's argument is the analogy with how humans learn temperature. You walk into a new room, you do not restart your model of heating from scratch. You reuse what you already learned in every other room. Federated learning gives the system the same property: every building contributes to a shared model of how to heat, without ever exposing the building's identifying data, and the heating logic is portable to the next deployment without retraining from zero.