FairAgent : a way towards FAIR Metadata with LLMs

This presentation introduces FairAgent, a schema-grounded system that supports the conversion of heterogeneous scientific imaging metadata into FAIR NeXus/HDF5 structures. The approach combines deterministic metadata mapping, NeXus schema constraints, selective LLM assistance, validation, and caching to make metadata harmonization more scalable, reproducible, and suitable for multi-technique research facilities.