Skip to content

Home

Overview

The Fusion Data Platform (FDP) helps researchers work with fusion experimental and simulation data, from raw signals through to published results. It includes TokSearch for building data pipelines, the Common Metadata Framework for provenance, an AI assistant that writes pipelines from plain-language requests, and a federated data layer that serves facility data without copying it. Together they support data reuse, provenance tracking, and reproducible workflows across different fusion experiments.

Why FDP Exists

Fusion research generates a lot of data across machines, diagnostics, and simulations, and most of it is hard to reuse. Getting at it means knowing which archive holds what, in which legacy format, behind which access mechanism. Reusing someone else's result means being able to tell what produced it.

FDP connects ingestion, curation, analysis, and sharing so that work happens once rather than in every group separately, and so that results carry enough metadata to satisfy FAIR (Findable, Accessible, Interoperable, Reusable).

Get Involved

We welcome participation from researchers, engineers, and AI/ML specialists.

  • Contribute workflows and datasets to the fusion AI/ML ecosystem.
  • Collaborate on AI-driven fusion research using FDP's tools.

The code is on GitHub at github.com/GA-FDP, and bug reports and feature requests are welcome as issues on the relevant repository.

📌 FDP currently provides efficient access to DIII-D and MAST/MAST-U data. DIII-D access requires being a DIII-D User. If you are not yet one, follow the steps here. MAST/MAST-U is served from the public FAIR MAST archive and needs no token.