The NVIDIA AI Technology Center (NVAITC) at UF is a joint research center of University of Florida and NVIDIA for advancing Artificial Intelligence (AI) education and research, as well as fostering partnership between higher education and industry. The NVAITC-UF is the first NVAITC in North America. The close collaboration enables UF researchers to adopt the latest technologies developed at NVIDIA and accelerate their projects to new levels. NVAITC-UF collaborations can be research group or individual researcher oriented on their projects. In addition, the NVIDIA Deep Learning Institute (DLI) training programs greatly enhance UF’s AI education from fundamental AI skills to advanced AI application development and deployment.

NVAITC Site Manager at UF: Dr. Kaleb Smith, Senior Data Scientist at NVIDIA

NVAITC Collaborations

The NVAITC is University wide, so a researcher from any college or department can become a NVAITC collaborator. Successful NVAITC projects are those with a clear vision and an execution plan. Typically, projects will have a team (a researcher and at least one student/staff) proficient in coding, an identified problem, and some solutions already explored.

How can UF researchers work with NVIDIA?

• Collaborations come in many forms, but the crux of this collaboration is NVIDIA scientist’s support/consultation at expert level for your research projects
• Acceleration of workflow pipeline, scaling model/workflow, performance improvement on models/algorithms, novel models/algorithms, consultation during the entire workflow is also possible with more guidance from NVAITC than hands-on coding/work.
• NVIDIA experts are not extra coders or software developers

What are the benefits of NVAITC collaboration?

• NVIDIA expertise in research area
• NVIDIA Expert’s consultation: brainstorming, compute, mathematical modeling in various scientific disciplines
• Industry/research look into solving hard problems
• Dissemination of research work
• NVIDIA blogs, GTC presentations, SDK packaging 

What are realistic expectations?

• It is a collaborative exercise
• Minimum monthly meetings on project updates. More frequent when needed to meet deadlines.
• Learning from all sides of the work with all parties working towards high caliber research to impact the community

Project Submission and Selection

If you are interested in collaborating with NVAITC on your research project, please follow the steps below:

1. Email Dr. Kaleb Smith (kasmith@nvidia.com) if interested in learning more about the NVAITC and if your project might be a good fit. 
• Dr. Smith and the UFIT Research Computing’s AI team manager Ying Zhang (yingz@ufl.edu) are available for consultation during the initial exploration process.
2. Fill out the project application form using the template provided by Dr. Smith after initial NVAITC consultation/project briefing.
• The proposal should include a clear statement of work (SOW) outlining the project participants and their roles and responsibilities.
3. The review of the proposal takes a couple of weeks. The evaluation criteria include
• NVAITC criteria:
• Targeted publication for the proposed work
• Software stack used in the the project
• Scale of the computations and resources needed for the projects. This goes both ways: some small scale projects have the advantage of fitting easily on the system for rapid completion; large scale projects are harder to schedule, but may be of interest because of potential larger scientific and technical impact.
• A identified team of students, postdocs, or research staff as well as the PI to work with on the collaboration. 
• Project characteristics: compact timeline, availability of external funding, potential for future funding,  etc.
• Details of the SOW.
• Realistic work being done from both sides, the project is identified with an obtainable outcome (time constraints is not a factor), data identified, team identified, and a timeline in place.

Acceptance and Notification

Once the proposal has been accepted, the PI will be contacted and the proposed SOW will be expanded and modified to become a full SOW for execution of the project.