Reliable AI and Data Optimisation

Projektinformationen

RAIDO

ID Finanzhilfevereinbarung: 101135800

DOI

10.3030/101135800

EK-Unterschriftsdatum 23 Oktober 2023

Startdatum 1 Januar 2024

Enddatum 31 Dezember 2026

Finanziert unter

• Digital, Industry and Space

Gesamtkosten

€ 7 738 182,50

EU-Beitrag

€ 7 738 182,50

Koordiniert durch

INSTITUT JOZEF STEFAN
Slovenia

Periodic Reporting for period 1 - RAIDO (Reliable AI and Data Optimisation)

Berichtszeitraum: 2024-01-01 bis 2025-06-30

Zusammenfassung vom Kontext und den Gesamtzielen des Projekts

Artificial Intelligence (AI) is transforming every sector of society and the economy, from energy and healthcare to agriculture and manufacturing. However, current AI systems often demand substantial computational resources, resulting in high energy consumption, increased costs, and environmental impact. At the same time, issues of fairness, trust, and compliance with European values and regulations are becoming increasingly urgent.
The RAIDO project (January 2024–December 2026) addresses these challenges by developing a framework for sustainable, trustworthy, and energy-efficient AI. RAIDO brings together leading European research and industry partners to create methods, tools, and governance frameworks that ensure AI systems are high-performing, environmentally responsible, legally compliant, and socially beneficial.
By optimizing AI workflows and datasets across both edge and cloud environments, RAIDO reduces computational waste, lowers operational costs, and enables reliable AI deployment in critical domains such as smart energy grids, healthcare, robotics, and precision farming. The project directly contributes to EU priorities on green digital transformation, human-centric AI, and trustworthy technologies, while laying the groundwork for scalable, ethical, and energy-aware AI solutions.

Arbeit, die ab Beginn des Projekts bis zum Ende des durch den Bericht erfassten Berichtszeitraums geleistet wurde, und die wichtigsten bis dahin erzielten Ergebnisse

During its first phase, RAIDO has delivered important scientific and technical results that set the foundation for pilot applications:
• System Architecture & Requirements (WP2 – completed): Designed the RAIDO system architecture, pilot requirements, and performance indicators. Delivered the RAIDO Green AI Framework and Optimised Pipeline, integrating monitoring tools (Prometheus, Grafana, InfluxDB, Kepler, MLFlow) to track both energy use and AI performance.
• Data and Model Optimisation (WP3): Built a robust data curation pipeline to address missing data, noise, and outliers; developed synthetic data generation methods using generative AI and digital twins; and implemented federated learning successfully tested on low-power devices such as Raspberry Pi 5.
• Legal and Ethical Frameworks (WP4): Conducted a comparative GDPR analysis in Belgium, France, Spain, and Greece. Explored privacy-preserving alternatives, including synthetic data and digital twins. Initiated the AI Ethics & Principles by Design Framework and developed early prototypes of an explainable AI (XAI) engine.
• Core Orchestrators & Interfaces (WP5): Advanced the RAIDO orchestrator for energy-aware AI workflows, created a blockchain-based monitoring system to ensure transparency, and developed a user-centric visualisation engine.

Fortschritte, die über den aktuellen Stand der Technik hinausgehen und voraussichtliche potenzielle Auswirkungen (einschließlich der bis dato erzielten sozioökonomischen Auswirkungen und weiter gefassten gesellschaftlichen Auswirkungen des Projekts)

RAIDO is pioneering new capabilities that go well beyond current AI practices, laying the groundwork for scientific breakthroughs, industrial uptake, and regulatory innovation:
• Energy-aware AI orchestration: The RAIDO Green AI framework enables real-time monitoring and optimisation of energy use during AI training and deployment — a functionality absent in mainstream platforms.
• Federated AI on constrained devices: Demonstrated distributed learning on low-power edge devices, unlocking decentralised, privacy-preserving AI applications.
• Ethics and transparency by design: Introduced explainability tools, trust levels for model auditing, and proactive integration of legal/ethical requirements, offering a replicable blueprint for human-centric AI systems.
• Synthetic data and digital twins: Developed GDPR-compliant data alternatives that accelerate innovation in sensitive domains such as healthcare, energy, and smart farming.
Looking forward, RAIDO will scale these advances through pilots, open-source tools, engagement with standardisation bodies, and collaboration with industrial partners to maximise uptake and market readiness.

RAIDO logo