Planning an NSF AI Datasets Proposal? Globus is Here to Help
September 30, 2026 | Lee Liming
NSF’s Unlocking Dataset Value for AI-Enabled Scientific Discovery program (NSF 26-512) closes November 4, 2026. It asks research communities to make datasets they already have discoverable, interoperable, governable, and ready for automated AI analysis — and encourages teams to build on existing national infrastructure rather than their own.
The Globus platform has robust cyberinfrastructure capabilities you can leverage for exactly this work, and we would like to help you put them to use. If you are preparing a proposal, we can help design the data architecture, show your community what it will look like before you submit, and support you as a named collaborator.
What Globus brings to the project
Discovery. Globus Search lets you create hosted, schema-flexible metadata indices built for scientific data: faceted and free-text query, arbitrary metadata models, per-record visibility controls. Sustained by Globus’s broad base of institutional subscribers, your indices are not dependent on ongoing project funding! Once your data catalog is loaded into an index, you can enable human discovery with Globus’s serverless web interface or a self-hosted Django portal, or enable LLMs to efficiently discover and analyze your data collection with an MCP Globus Search tool or Skill.
AI-ready pipelines. Globus Flows and Globus Compute run feature extraction, metadata generation, QC, harmonization, and embedding jobs next to the data, on campus clusters, national lab systems, or cloud, writing results back into the index. Triggered by events or on a schedule, curation becomes an ongoing service rather than a one-time conversion. Access and governance. Globus Transfer, HTTPS access, and connectors spanning various storage systems (e.g. POSIX, tape, AWS S3, Google Drive/Cloud, Azure Blob/OneDrive and many more) let a dataset stay federated across the institutions that hold it while presenting it as a single collection. Globus Auth, Groups, and record-level authorization give you federated login and real access management, with audit logging and High Assurance and BAA tiers for protected data. Guest collections and submission flows give your community a governed way to contribute.
Repositories already built on Globus
You do not have to imagine this in production. Major community data resources already run on these services:
- Federated Research Data Repository (FRDR) — Canada’s national multidisciplinary repository for large-scale deposit, discovery, and distribution.
- Materials Data Facility — materials science datasets published with programmatic access for machine learning.
- Earth System Grid Federation — distributed access to climate model output across international centers.
You may already have Globus
Most U.S. research universities, national laboratories, and major facilities already hold Globus subscriptions covering capabilities an AI Datasets project needs, such as index, portal, pipelines, transfer and sharing, for everyone at the institution. In many cases the license cost to your project is zero, so start with your research computing group rather than a budget line; ask us and we will confirm your coverage. Non-profits under $4M in annual research budget now qualify for a free Standard subscription.
How we work with proposal teams
- Architecture design sessions, at no cost. A session with our solutions architects to turn your dataset, community, and AI goals into a concrete design you can cost and describe.
- A prototype you can point to. We can help stand up a pilot index, portal, and flow with real data before the deadline, so the proposal describes something that exists.
- Letters of collaboration, prepared to NSF’s requirements. After award we stay involved: deployment support, tutorials for your community, and a sustainable not-for-profit platform from the University of Chicago that runs long after the grant ends.
Talk to us
Tell us about your dataset and your community and we will set up a call — ideally six to eight weeks ahead of the deadline.