Imagine if building an AI was like choosing a pre-made meal instead of cooking everything from scratch every time. This tool helps developers pick the most energy-efficient AI models by recycling existing knowledge and matching the software to the hardware. It stops the waste of electricity and computing power that usually happens during the trial-and-error phase of AI development.
AI development currently prioritizes speed and accuracy, leading to massive energy waste and high CO2 emissions. This creates a conflict between the need for AI competitiveness and environmental sustainability goals.
A carbon footprint based model optimization tool and a knowledge graph database for recycling AI functional cores.
If you are a cloud provider dealing with skyrocketing electricity bills for AI training — this project developed a carbon footprint based model optimization tool that reduces the energy requirements of AI systems. This allows you to offer greener, lower-cost computing services to your clients.
If you are a hardware maker dealing with limited battery life in smart devices — this project developed a co-design tool that matches AI models to specific hardware. This ensures AI features run efficiently without draining the device battery.
If you are a software firm dealing with long, expensive AI experimentation cycles — this project developed a library of functional knowledge cores that allows for recycling AI components. This speeds up development and cuts down on wasted computation epochs.
Based on available project data, specific pricing or licensing costs are not mentioned, though the project received an EU contribution of EUR 3,742,855 for development.
Yes, the project specifically targets the final training of production systems and continuous online re-training during deployment to ensure industrial scalability.
Based on available project data, the results will be made available on the AI4EU platform, but specific IP ownership details are not provided.
The project addresses the fact that current tools like PyTorch prioritize performance over energy, aiming to provide a toolchain that integrates with these workflows to add sustainability metrics.
The project period runs from 2022-10-01 to 2026-06-30, indicating it is currently in the development and validation phase.
The consortium is well-balanced for commercialization, featuring a 43% industry ratio with 3 industrial partners and 2 SMEs. The presence of partners from 5 countries (CH, DE, DK, ES, FR) suggests a strong European market reach, combining academic research from 2 universities and 2 research centers with practical industrial application.
Contact PROYECTOS Y SISTEMAS DE MANTENIMIENTO SL in Spain
Contact us to connect with the SustainML consortium for early access to the optimization tool.
