Reliable AI and Data Optimisation
The EU-funded RAIDO project aims to deliver a unified framework for the development of trustworthy and energy-efficient artificial intelligence (AI), addressing the growing need for reliable, transparent, and sustainable AI systems across critical sectors. As AI technologies become increasingly integrated into domains such as healthcare, smart farming, energy systems, and robotics, ensuring their safety, fairness, and environmental sustainability is essential. RAIDO tackles these challenges by combining advanced methodologies for both Trustworthy AI and Green AI, enabling the development of systems that are not only accurate and reliable, but also resource-efficient and environmentally responsible. The project introduces automated data curation and enrichment techniques, leveraging tools such as digital twins and diffusion models to generate high-quality, unbiased, and compliant training datasets. In parallel, RAIDO incorporates data- and compute-efficient approaches, including few- and zero-shot learning, model optimisation, and continual learning, to significantly reduce the energy footprint of AI development and deployment.
To enhance transparency and trust, RAIDO integrates explainable AI (XAI) techniques, blockchain-based mechanisms, and advanced monitoring and evaluation methods. These components ensure that AI models and data processes remain interpretable, auditable, and aligned with ethical and regulatory requirements. At the core of the framework lies an intelligent AI orchestrator, which coordinates data processing, model training, and deployment workflows, optimising performance while minimising energy consumption. The RAIDO framework will be validated through four real-world demonstrators across key application domains, including smart grids, computer vision-based farming, healthcare, and robotics, highlighting its potential to deliver tangible societal and industrial impact.
