BU AI Research Resource Program (BU-AIRR)

AI research increasingly requires large amounts of computational resources (in the form of Graphics Processing Units – GPUs) that in their most advanced form have a cost that is beyond the reach of individual research groups. As a result, much of the advanced AI research to train large deep neural network models is performed on the cloud. The Mass Open Cloud Alliance, based at the Hariri Institute for Computing and Computational Science & Engineering, is offering cloud services for research through the New England Research Cloud (NERC). NERC is offering access to top-of-the-line CPUs, storage solutions, a number of Nvidia A100 GPUs, and has recently acquired 192 Nvidia H100 GPUs that are expected to go on-line in December 2024. NERC is a fee/hour service available to any BU researcher, and researchers from several other institutions.

In an effort to enable and seed large-scale AI research, Boston University is establishing a program to provide credits for compute time on these high-end GPU resources. A pool of funds has been established through contributions from the Hariri Institute, the Red Hat Collaboratory based at the Institute, the Department of Electrical & Computer Engineering, the College of Engineering, and the Office of Research.

BU researchers who would like to use these resources are encouraged to submit a short proposal outlining their plans. The initial deadline is November 16, 2024, with rolling deadlines thereafter and for as long as funds are available.

For any questions, please reach out to Nancy Clinton, at nclinton@bu.edu.

Proposal Requirements

Proposers should apply by filling out the web form below. Researchers can either upload a PDF or select the “Proposal Form” option to provide responses to prompts directly in the web form. Proposals are due November 16, 2024, with rolling acceptance of proposals after the due date, if funding permits.

The length of the proposal should be around one page and should not exceed two pages using font and formatting instructions applicable to NSF proposals. References are excluded from the page length limit.

The proposal should include the following components:

1. Scientific Goal and Potential Impact

Please explain the scientific goal and potential impact of your project.

2. Estimate of GPU Compute and Storage

Please provide detailed information about the number of experiments, what each experiment will entail, and an estimated number of NVIDIA A100 or H100 GPU hours and the specific cost associated with those hours. Please “show your work” in coming up with estimates. NERC pricing can be found here.

3. Sponsored Programs that will be Enabled

This can include current projects and future proposed projects. Provide specifics about current projects (PIs, title, sponsor, award number, dates, and amount) and future proposed projects (program, sponsor, agency, deadlines).