Advancing Machine Learning and Interoperability for Climate Adaptation: Highlights from the Climate-Adapt4EOSC & FAIR2Adapt Joint Workshop

Climate-Adapt4EOSC hosted a joint online workshop with sister project FAIR2Adapt on 31 October, focused on advancing machine-learning standardisation, FAIR data interoperability, and international collaboration to support climate adaptation research within the European Open Science Cloud (EOSC) framework. The session centred on a presentation by Vyacheslav (Slava) Tykhonov, Head of Interoperability & AI at CODATA and one of Climate-Adapt4EOSC’s senior Data Scientists, who introduced the latest developments in the Croissant Standard for Machine Learning and the Cross-Domain Interoperability Framework (CDIF).

Croissant Standard for Machine Learning

Slava introduced the emerging Croissant Standard, designed to structure and document machine-learning workflows in a transparent, reusable, and interoperable way. The standard supports reproducible AI processes and promotes the sharing of adaptation models across research projects and infrastructures.

CDIF – Cross-Domain Interoperability Framework

He outlined the CDIF framework, which enables the integration of heterogeneous datasets — from climate and hydrology to socio-economic domains — ensuring they are FAIR and machine-actionable. This work directly supports the European Open Science Cloud (EOSC) vision of seamless, cross-disciplinary data reuse.

International Collaboration with Chile

Slava also highlighted ongoing cooperation with a partner team in Chile, where the CroissantML and CDIF approaches are being validated using diverse environmental dataset, a demonstration of the adaptability of the standard internationally.

Significance for the Climate-Adapt4EOSC Project

For Climate-Adapt4EOSC, the workshop marked an important step forward in its work on FAIR Digital Objects, ontology development, and semantic interoperability. Following the presentation, a lively Q&A focused on CroissantML and CDIF and on how the project’s outputs can bridge research data infrastructures and AI-driven analytics — transforming FAIR data into applications for climate adaptation.

The workshop was also an opportunity to further strengthen collaboration with FAIR2Adapt and explore how the outcomes of both projects can be integrated within the broader EOSC framework.

About Vyacheslav Tykhonov

With over 25 years’ experience in software development, data science, and European research infrastructures, Slava is an innovator, IT architect, and mentor whose has focused much of his career on research and education through experimentation, pilot projects, and inventive problem-solving methods. He is a Harvard Dataverse Ambassador (since June 2025) and co-author of the CroissantML standard for AI.

Slava has contributed to numerous open-source and interdisciplinary research initiatives, serving as technical lead and advisor in projects spanning semantic web technologies, cloud deployment, NLP and ML applications, digital humanities, and big-data infrastructure. Beyond his FAIR-data and AI work, he sits on the boards of the non-profit organisations CoronaWhy and the COVID-19 Museum, contributing to research that supports global health preparedness and public understanding of pandemics.