The work planned in the Turing project will be carried out in eight (8) work packages (WPs).

WP 1 | Project Coordination

WP1 ensures the effective coordination, administration, and overall management of the TURING project. It establishes the administrative and financial framework, supervises the scientific and technical progress, and enforces quality assurance across all activities. The work package also promotes gender equality and ethical compliance, manages risks and opportunities, and ensures smooth communication among partners. In addition, it drives innovation management and open science practices while safeguarding adherence to data protection, privacy, and AI ethics principles through continuous ethical oversight and an independent Ethics Advisory Board.

WP 2 | Research scope specification

WP2 establishes the scientific and technical foundation of the TURING project, defining the framework and requirements that guide all subsequent developments. It involves the analysis of complex dynamical systems, existing foundation models, and available datasets across domains such as nuclear energy, high-energy physics, and meteorology. The work package defines the data requirements, curates and prepares datasets, and specifies the architecture and functional design of the TURING framework. Emphasis is placed on model robustness, explainability, privacy, and compliance with legal and ethical standards, ensuring a solid, trustworthy basis for AI-driven scientific research.

WP 3 | Pre-training TURING models

WP3 focuses on developing and pre-training the foundation models that underpin the TURING framework. The work package integrates physics-informed AI approaches to ensure models remain scientifically consistent while achieving computational efficiency and interpretability. It advances multi-application meta-learning and federated training techniques to support diverse domains, while improving model robustness, privacy, and security through adversarial training and distributed transformer architectures. Overall, WP3 establishes the foundation for scalable, trustworthy, and physically grounded AI models capable of adapting to complex scientific phenomena.

WP 4 | Finetuning TURING models

WP4 aims to adapt the pretrained foundation models to specific scientific and industrial applications, transforming them into fully operational TURING models. It develops few-shot and low-shot learning methods to efficiently fine-tune large models while maintaining generalization and preventing overfitting. The work package also advances uncertainty quantification and explainable AI techniques to ensure trustworthy and transparent model behavior. In parallel, it optimizes inference efficiency through model compression, pruning, and dynamic architectures, enabling high performance and adaptability across diverse computational environments.

WP 5 | TURING framework

WP5 delivers the core offerings of TURING to the R&D community and practitioners. It provides the computational resources and user-centric framework that enable the development, deployment, and interaction with TURING models. The work package sets up the distributed computing infrastructure and scalable storage systems, ensuring secure and efficient access for model training and inference. It implements the TURING Framework, a web-based interface with conversational and visual interaction capabilities, enabling seamless user engagement with the models. Moreover, WP5 develops standardized MLOps pipelines and automation workflows to streamline model training, fine-tuning, deployment, and integration. The result is a robust, scalable, and accessible framework that supports the entire lifecycle of TURING models.

WP 6 | Experiment realization in targeted application domains

WP6 validates and demonstrates the TURING models across the targeted application domains. It defines standardized experimental protocols to ensure consistency, reproducibility, and compliance with trustworthiness, privacy, and security standards. The work package executes experiments in real-world and simulated environments, assessing model performance, robustness, scalability, and usability. It also conducts user-centric evaluations to measure trustworthiness, key performance indicators, and overall impact, enabling continuous refinement of the TURING framework and ensuring its practical applicability for scientific research and operational decision-making.

WP 7 | Outreach, impact, and social inclusion

WP7 focuses on maximizing the visibility, societal impact, and inclusiveness of the TURING project. It develops and implements communication and dissemination strategies to reach target communities while promoting accessibility and diversity. The work package ensures alignment with EU standards and evolving AI regulations and establishes collaborations with other EU-funded projects, initiatives, and international partners. WP7 also drives exploitation and sustainability activities, fostering long-term application of project results in scientific, industrial, and societal contexts, and supports capacity-building through workshops, secondments, and PhD positions.

WP 8 | Ethics requirements

WP8 ensures that all project activities comply with established ethics requirements. It monitors adherence to ethical standards across the TURING project, including privacy, data protection, and responsible AI practices, guaranteeing that research and development are conducted in line with European and international norms.