AI/ML Workflow Orchestration and E2E Energy Footprint Computation

Extensive research expertise with development of intelligent orchestration and workflows for AI/ML, including open source software and experimentation in trials.

NAOMI — the only available alternative to the O-RAN AI/ML workflow. Designed from the ground up for modularity and distributed training, NAOMI empowers AI/ML model execution across heterogeneous edge devices — not as a workaround, but by design. If you're building next-gen intelligent RAN systems, NAOMI is the framework to watch: https://github.com/sensorlab/naomi

eCAL - a new metric for computing the energy consumption throughout the architectural components and lifecycle of an AI-powered wireless systems by analyzing the complexity of data collection and manipulation in individual components and deriving overall and per-bit energy consumption: https://github.com/sensorlab/eCAL

MRM3: Machine Readable ML Model Metadata - better describing machine learning models by including energy and hardware information, in addition to data and performance. Such information enables better orchestration for training and inference tasks: https://github.com/sensorlab/MRM3