The National AI-Ready Research Data and Computing Model

Need and rationale

Federal research policy now expects every institution that receives federal funds to plan, secure, document, and share its research data. The NSF Proposal and Award Policies and Procedures Guide and the NIH Data Management and Sharing Policy apply the same requirements to a small college with a few funded investigators as to a large research university.

Federal agencies also provide computing for data-intensive and AI research at no cost to the institution. The Open Science Pool (OSPool), operated by the OSG Consortium with support from the NSF-funded Partnership to Advance Throughput Computing (PATh), provides high-throughput computing. NSF ACCESS allocates time on national supercomputers. The National Artificial Intelligence Research Resource (NAIRR) Pilot provides computing, data, and models by application.

The staff who turn these requirements and resources into daily practice are concentrated at a small number of institutions. Under the Carnegie Classification's 2025 Research Activity Designations, 187 institutions are R1 and 139 are R2. The National Center for Education Statistics counted 2,637 four-year degree-granting institutions in 2020-21. Roughly 2,300 four-year institutions, and every two-year college, sit outside the tier that research computing and research data services are usually built around.

At these institutions I find three recurring gaps: no research computing staff to prepare allocation requests or onboard researchers; no local cluster, and therefore no local expertise; and no research data office to help faculty meet data management and research security requirements. The result is lower competitiveness for proposals that require data management and computing plans, difficulty recruiting and retaining computationally trained faculty, and students who graduate without experience in reproducible data work.

Definition

In the NARDC model, an institution designates or trains one person with a defined role instead of acquiring a cluster. That person translates federal requirements into local templates, trains faculty and students, and connects researchers to national cyberinfrastructure that the federal government already funds. The facilitator works from a documented set of open materials and remains engaged until the institution can sustain the work with its own personnel.

Scope

The model does not involve acquiring a computing cluster, establishing a research computing department, or delivering a single training event. It is not tied to one position at one institution; it is a method that an existing position in a library, IT unit, or research office can carry.

Intended institutions

R2 universities, Research Colleges and Universities, minority-serving institutions, primarily undergraduate institutions, and community colleges, in particular those where research data and computing support rests with one person or has no designated owner.

Model design

The model has three components, delivered through a six-stage cycle that repeats as new departments and research needs are brought in.

Components

• Policy translation. Federal data management, data sharing, and research security requirements converted into templates, workflows, and training. The core instrument is a campus research data management plan template covering project summary, ownership and access, sharing restrictions and licensing, funding and affiliation, roles and access privileges, storage, security, and preservation.
• Workforce development. Training for faculty, staff, and students in secure, reproducible, and AI-ready data practice, using the tools already in use (R, Python, Stata, SPSS). This includes short-format workshops, redesigned credit-bearing courses, and a student research incubator in which trained students contribute to faculty research projects.
• Shared cyberinfrastructure. Researchers are connected to national resources, including the OSPool, ACCESS, and the NAIRR Pilot, in coordination with each program's facilitation staff, in place of local hardware acquisition.

Six-stage cycle

The cycle is iterative. Support informs subsequent rounds of Assess and Design, and different departments may be at different stages at the same time.

Criteria for self-sufficiency

The model is evaluated by whether the institution can sustain it without the original facilitator. An institution reaches the Release stage when the following conditions hold for two consecutive semesters:

• at least one local faculty or staff member delivers training without outside assistance;
• researchers submit and complete jobs on national platforms independently;
• the data management plan template is in routine use for proposal submissions;
• the student incubator continues to complete faculty research projects; and
• templates, guides, and workshop materials are archived and maintained by the institution.

Resource requirements

The fixed cost is one trained person. Many institutions already have a position in the library, IT, or the research office that can assume the role. Computing is provided by federally funded programs at no charge to the institution, and the materials are openly licensed for reuse.

Pilot implementation: Seton Hall University, 2020 to 2026

When I joined Research Data Services at Seton Hall University Libraries in February 2020, the university had no research computing cluster and no research computing staff. I developed and tested the model there. The figures below are drawn from documented records (annual reports, LibCal and LibAnswers exports, course evaluations, and program logs). Where two sources differed, I report the lower figure.

Activities by component

Policy translation. I developed the Seton Hall University Libraries Research Data Management Plan Template (first version 2022; current campus-wide version 2026), a single fillable document aligned with federal data management expectations.

Workforce development. In 2022 I co-founded the DataLab, a student research incubator that places trained students on faculty research projects. I redesigned three credit-bearing courses around reproducible data practice and computing: health sciences statistics, geographic information systems, and doctoral statistical methods. I added AI-readiness training to the program, including an AI Literacy workshop series and, in March 2026, "Building Private AI Over Institutional Data" and "Agentic AI in Higher Education." Love Data Week, which I led in 2025 and 2026, drew participants from other institutions and organizations each year.

Shared cyberinfrastructure. I initiated Seton Hall's first researcher onboarding to the NSF-funded OSPool and remain an active user. I wrote an OSPool login guide for faculty without command-line experience and, in 2025, held onboarding consultations with OSG facilitation staff and the Center for High Throughput Computing at the University of Wisconsin-Madison. A DataLab student used the OSPool for a project that received a student research award at the university's academic exposition.

Outcomes and limitations

With one position and no cluster, the institution established a data management plan template, a recurring training program, a student incubator that completes research projects, and a pathway to national computing. Some faculty trained through the program now deliver training to colleagues, which marks the beginning of the Release stage.

The model has been implemented at one institution, under my direction, so the pilot does not yet show how it performs with other personnel and in other settings. Most indicators are counts of services delivered rather than research outcomes such as awards or publications. Implementation and evaluation at additional institutions are designed to address these limitations.

Relation to existing approaches and programs

Institutions seeking research computing capacity typically consider three approaches. The table compares them with the NARDC model for an institution with no existing capacity.

Complementary national programs

National programs, among them OSG School, the Virtual Residency, CI Pathways, CyberAmbassadors, and the ACCESS campus champions network, train research computing facilitators and users. I have participated in several of them, and the NARDC model depends on them. These programs train individuals and generally assume that participants return to institutions with infrastructure, colleagues, and an institutional mandate. At limited-resource institutions those conditions are usually absent.

The NARDC model addresses the period after training: what to assess first, which templates to adapt, how to sequence instruction, how to bring the first researchers onto national computing, and how to support them through early difficulties. It also combines three functions that R1 universities generally distribute across separate units (research data services, research computing facilitation, and student research training), because at a smaller institution these functions typically fall to one person.

Institutional partnership

Institutions may adopt the model independently using the open materials or in partnership with me through one or more stages of the cycle. Each engagement begins with an initial consultation on the institution's research profile, existing resources, and priorities.

Institutional commitments

• A sponsoring unit (library, IT, or research office) accountable for outcomes.
• One designated staff member with protected time for the role.
• Access to faculty in two or three departments with data-intensive or computational research.
• Documentation of workshops, consultations, onboarded researchers, and completed projects from the start of the engagement.

Evaluation

For each participating institution I record the same measures at baseline, after one year, and after two years.

Results are published annually with the materials, so that institutions and funders can assess what the model achieves and where it falls short.

Implementation guidance

This guidance is written for a library, IT unit, or research office implementing the model with no existing research computing capacity.

Staffing

Designate one person, full time or with protected time, as the institutional contact for research data and computing. Relevant preparation includes working knowledge of at least one statistical or programming language, familiarity with federal data management requirements, and willingness to learn the national platforms through their no-cost training. Place the role where faculty already seek assistance. At Seton Hall, the role sits in the University Libraries.

First six months

1. Meet with faculty in departments with data-intensive or computational research. Record research areas, data holdings, tools in use, and applicable compliance requirements.
2. Adapt an open data management plan template to institutional policies and offices.
3. Open an OSPool account and complete its onboarding; review ACCESS and the NAIRR Pilot. Onboarding others is more effective once the facilitator is an active user.
4. Deliver an initial series of four to six short-format workshops in the tools in local use, including one on version control and reproducible workflows.
5. Select one faculty member with a defined computing need and bring that research group onto a national resource with the program's facilitation staff.

First two years

1. Establish a student research incubator that pairs trained students with faculty projects through application, matching, supervised work, completion, and presentation.
2. Redesign one existing credit-bearing course to include reproducible workflows and national computing.
3. Document consultations, workshops, onboarded researchers, and completed projects from the outset. Funders, administrators, and peer institutions will request these records.
4. Invite faculty who have completed onboarding to deliver training to colleagues.

Guiding principles

• Use federally provided resources before acquiring local infrastructure.
• Teach in the tools researchers already use.
• Report only what is documented.
• Plan for follow-up support after training.
• Reduce the institution's reliance on outside assistance each year.

Related resources

• National research computing for small institutions
• OSPool or ACCESS? A guide to national computing
• Your first job on a research computing cluster
• Starting research data services at a small institution
• Writing a data management plan
• Research computing readiness checklist

Frequently asked questions