| Abstract Scope |
Unlocking the full potential of AI in materials science requires moving beyond static archiving to dynamic, structured data management. We present a platform that integrates FAIR principles through a semantic knowledge graph, transforming experimental workflows into machine-actionable, interoperable datasets. By capturing rich provenance at the source using visual, semantically typed workflows, we eliminate the "data wrangling" bottleneck that plagues current AI initiatives. As a use case, we focus on tribology, where material performance relies on a high-dimensional context.
We demonstrate how this ontology-driven approach supports a community-wide FAIR portal, utilizing persistent identifiers to facilitate collaboration across diverse labs. This infrastructure harmonizes complex, multi-modal data, ensuring it is ready for AI applications, including data fusion and quality enhancement. Ultimately, we provide a blueprint for turning scattered experimental results into trustworthy, high-quality training data that drives reproducibility and accelerates materials discovery. |