The AIxpertise joint project aims to build a structured, collaborative platform for experts to harness Artificial Intelligence (AI) for the nuclear energy sector with robustness and transparency. Project participants will engage in hands-on activities in three pillars: data, benchmarking, and training. They will enhance knowledge, collect and curate data, and assess tools to apply AI solutions effectively and safely for the future of the nuclear energy sector.
Artificial intelligence (AI) and machine learning (ML) applications are poised to revolutionise nuclear research and development. Potential high-impact benefits of AI and ML applications include autonomous plant operation, generative AI-based expert systems, advanced nuclear reactor design, and revolutions in scientific modelling and simulation. These breakthroughs are enabled by the seamless integration of experimental and operational data with physics-informed and self-learning models.
The AIxpertise proposal outlines an opportunity for partners to establish a joint project within the NEA to harness the transformative potential of AI. By establishing a collaborative platform focused on data, benchmarking and training, the project will drive innovation, expand expertise across various research areas, and ensure the sector remains at the forefront of technological advancements. This proposal will foster transparency, trust, and support regulatory readiness for AI applications, leveraging decades of experience at NEA, its robust infrastructure and international reach.
Our vision is to build a future where AI facilitates and speeds up nuclear engineering and research processes, strengthens data-driven decision making, and drives progress towards safe and sustainable nuclear energy solutions.
The nuclear sector needs to keep pace with fast-advancing technological innovations in AI/ML and ensure readiness for regulatory review of AI approaches in a variety of applications. For experts in nuclear science and engineering, there must be a platform to investigate the benefits and opportunities of AI, to demystify AI solutions, to expand capacity, and to harness AI/ML in a clear and defensible way. This project proposal addresses that need by establishing a community-of-practice ready to learn by doing, with specific, structured deliverables in three project pillars highlighted in the table below.
| DATA Acquiring and curating datasets for AI applications. |
BENCHMARKING Assessing and training AI algorithms. Building bridges between nuclear and AI domains. |
TRAINING Leveraging project outcomes to develop education and training programmes for nuclear energy experts. |
| Precursor NEA Activities |
||
| 20+ years of experience in collecting, preserving and disseminating experimental data in different domains. |
Task Force on Artificial Intelligence and Machine Learning for Scientific Computing in Nuclear Engineering. |
20+ years of experience training in various fields, online and in person for professionals, researchers and university and high school students. |
| Project Activities and Outputs |
||
| AI-friendly internationally peer-reviewed, aggregated, and accessible datasets. |
International benchmarks for assessing and training AI algorithms. |
Trainings on the use of AI algorithm for nuclear research and education. |
| Build upon the work of the NEA expert groups that have successfully evaluated and preserved essential experimental data across various technical fields. Integrate data repositories across various research areas including time series data for testing predictive maintenance towards safety applications. Make the datasets machine-readable with application programming interfaces (APIs) and comprehensive user documentation. Conduct data mining to provide new datasets and flag inconsistencies in existing datasets. Perform target accuracy assessments and promote smart experiment designs to fill data gaps. Extend and implement continuous testing and improvement strategies for datasets, incorporating a robust feedback mechanism. |
Facilitate collaboration among nuclear scientists, engineers, and AI experts to address the needs of nuclear research and education communities. Demonstrate the performance and assess the explainability and robustness of AI algorithms through international benchmark activities. Leverage supervised and unsupervised ML, VVUQ, anomaly detection, generative AI, optimisation. Evaluate the effectiveness of Large Language Models and Semantic Search in streamlining access to the nuclear research knowledge base. Assess and train novel AI algorithms using well characterised NEA datasets and well-defined international benchmark exercises. Develop leaderboards, document best practices and lessons learnt from the benchmark exercises to validate AI-driven scientific computing. |
Offer resources for training and education including hands-on exercises based on data repositories to assess and train AI models. Organise AI/ML schools tailored for nuclear engineers, scientists and students to foster skill development. Develop and implement certified education programme in partnership with the NEA Global Forum on Nuclear Education, Science, Technology and Policy. Organise hackathon and ideation events to investigate the opportunities for applying AI algorithms and hardware for scientific computing. |
| Online platform for collaboration of nuclear and AI experts including GitLab repositories for aggregated data, benchmark specifications, software, AI models, and interactive discussions and issue boards. |
||
AIxpertise aims to build an online platform for collaboration of nuclear and AI experts to produce specific, structured deliverables in three project pillars in an initial three-year scope of work by leveraging existing data assets, members’ in-kind contributions as well as technical support contracts for member-defined scope. The access-managed secure platform will include GitLab repositories for aggregated data, benchmark specifications, software, AI models, and interactive discussions and issue boards (included into the GitLab offering). Project funding would ensure, among others, meaningful work and outputs in maintaining the platform, curating and annotating machine-readable datasets, and exercising AI/ML processes. The project aims to foster the development of AI expertise within the nuclear research, academia, and technical support organisations (TSOs).
This project seeks to establish a platform that brings together scientists and engineers with expertise in nuclear science and AI, fostering collaboration and bridging the gap between these diverse technical domains. The target audience includes research organisations, education institutions, safety authorities, TSOs, and industry.
The first phase of the AIxpertise project will span three years, with development of the project scope ongoing through 2025 in collaboration with interested parties. The project is expected to officially start in Q1 2026 and its first phase to progress until the end of 2028.
| March 2025: Perform global survey on needs and use cases. |
Q2 2025: Bilateral discussions to refine use cases. | Week 22-27 Sep 2025: AIxpertise Workshop - Define project scope. |
Q4 2025: Finalization of AIxpertise Joint Project Agreement |
Q1 2026: AIxpertise Project Launch
|
The NEA is looking forward to welcoming additional participants throughout the project development phase and interested parties are invited to contact the NEA for further information on the joining process. Please contact the NEA Secretariat at: AIXPERTISE@oecd-nea.org.
The project will establish a collaborative platform to bring together stakeholders from research organisations, safety authorities, technical support organisations (TSOs), industry, academia, and technology companies.
To be decided by participants.
OLIVER BUSS
Nuclear science
