ML Inference Engine for Smart Energy-aware Zero-touch Traffic Engineering

ML inference engine for Smart Energy-aware Zero-touch Traffic Engineering. Predicts router power consumption from instantaneous network telemetry-derived metrics to support Network Digital Twin environments and energy-aware traffic engineering. Includes a Dockerized inference service using Kafka topics for input/output and an optional telemetry mock service for replaying datasets, validating results, and generating plots. Supports two distinct carrier-grade router model families with linear regression, polynomial regression, random forest, and deep neural network variants. This work was carried out under the TC3.2 use case of the ACROSS European project (Horizon Europe programme)