Publish, curate, and discover data

Practical guidance for preparing submissions, curator review, DOI release, and reuse across the materials science community.

Share your data

MDF handles transfer, curation, DOI minting, and long-term hosting. The publishing process takes three steps.

01
Sign in with Globus
Use institutional or Google credentials through Globus. Most researchers do not need to create a separate identity.
Get started with Globus
02
Prepare your dataset
Organize data in durable formats such as CSV, HDF5, JSON, or NetCDF. Include a title, description, and at least one author.
Read dataset best practices
03
Submit and get a DOI
Complete the metadata and transfer details. MDF curators review the submission and mint a DataCite DOI.
Go to publish

01
Publishing from local storage
Upload from your computer, complete the metadata form, and submit for curation and DOI minting.
Coming soon
02
Publishing from a Globus endpoint
Link an existing Globus collection, capture its endpoint path, and submit without another file transfer.
Coming soon

Curators check metadata quality, data accessibility, citation readiness, and release status before a dataset becomes discoverable.

MDF datasets are freely accessible through the web, command line, or Python. Choose the workflow that fits your research.

01
Web interface
Search by keyword, organization, domain, or year, then review metadata, files, citations, and version history.
Dataset pages include complete metadata, Globus access, citation exports, version history, and a DOI.
Search datasets
02
MDF CLI
Authenticate once with Globus and clone a dataset by its source_id from the command line.
pip install mdf-cli · mdf login · mdf clone <source_id>
MDF CLI on GitHub
03
Foundry-ML
Load ML-ready datasets directly into Python as defined train and test splits without manual data wrangling.
pip install foundry-ml
Foundry-ML documentation