I build materials databases and the methods that make them trustworthy. For twenty-five years at the National Institute for Materials Science (NIMS), I have worked on collecting, structuring and validating materials data — from thermophysical properties of composites, thin films and interfaces, to the AWB battery materials database and the AI-assisted curation workflow behind it.
My current interest is a simple question with a long answer: how much of an expert curator's judgement can be written down, handed to AI, and measured? Scientific Data Works is where that question becomes practice — data collection, curation and analysis for materials research, done by one scientist and a team of AI agents.
Two decades of database design at NIMS, from AWA to the AWB battery materials database: schemas that keep the sample, the process and the measurement together with the number.
The BAI workflow: literature screening, extraction, normalisation and audit as separate AI roles under expert-defined rules — benchmarked against human curators, not assumed.
Thermal conductivity and interfacial thermal resistance of real materials. An open ITR dataset published in Scientific Data (2020); a condition-complete second version is next.