An artificial intelligence-powered platform that connects faculty researchers with new funding opportunities matching their research interests.

Developed in-house, Be$tMatch uses large language models to analyze NIH Notices of Funding Opportunity (NOFOs) published within two weeks and compare them with descriptions of research interests provided by faculty researchers at the HSC. The platform delivers the search outcome every two weeks. If matching NOFOs are identified for a faculty researcher, he or she will receive a personalized email notifications to look into these NOFOs more closely and make informed and timely decisions on whether and when to start developing grant applications.

Figure 1 shows a sample of investigators’ similarity score distributions for their top results. We can see that even without specific training, our off-the-shelf model is able to establish clear scoring hierarchies and, in many cases, human-intelligible groupings. While this supports our objective of identifying the most relevant funding opportunities, it is critical to acknowledge that score ranges differ significantly between investigators. This is attributed to variance in domains of research, as well as word choice in profile writing - since semantic similarity is evaluated across the full text of an investigator profile and FOA purpose statement, broad research domains and generic terms (e.g. “cancer”, “healthcare outcomes”) increase overall degree of similarity even when specific aims differ. While this can be affected by profile revision, this behavior is inherent to semantic similarity comparison and this points to the necessity of dynamically calculated or user-specific similarity score cutoffs.

Submitted a narrative but want to revise to receive more customized funding notifications for your research? Email Atsuko Bealmear.

Want to enroll in the Be$t Match program? Email Nicole Renner

Research Forrest