The availability of precise and comprehensive experimental data in science and technology is crucial for the usability of Artificial Intelligence (AI) models. To enable the deployment of data-driven applications across different platforms, a digitally analysable, system-independent representation of datasets is essential. We propose a metadata model based on domain-specific languages and terminologies, which allows researchers to focus on data provision by reducing routine activities rather than attempting to align with other research groups. Furthermore, it enables a fast and efficient integration of new partners from different laboratories and different disciplines. To conclude, our approach supports a paradigm shift away from more or less subjectively designed individualistic conceptions in handling research data towards objectively established harmonised solutions. The approach is illustrated for an Interdisciplinary Research Training Group, in which researchers from more than 10 different departments are involved with main research profiles, such as textile and polymer technology and material sciences.