Inspiration
With the new presidential administration, has come frightening uncertainties in sources of funding for research and other academic projects. The current administration has proposed dramatic cuts to the NIH and other government entities that provide critical sources of funding to academic institutions. Additionally, funding opportunities at Stanford and other universities are currently very decentralized across hundreds of web pages and email lists.
To address these challenges of funding and decentralization, we built Grants Search, a tool that helps University members find other funding opportunities within the University that are not directly dependent on government resources.
Also, in such a large scale research institution, it can be challenging and time-consuming to find faculty leading research that students and postdocs are most interested in. To address this, we also included actively funded NIH projects at the University in our database.
What it does
The user inputs a query of a project that they hope to find funding for or work on. Grants Search then searches for funding opportunities (grants, fellowships, scholarships), ongoing projects and research (including NIH funded work), and student and university organizations to find opportunities most closely related to the user’s project of interest. The funding amount and timeline is displayed for all opportunities. The user’s search can be further narrowed down by their academic status at the university (Undergraduate, Masters, Coterm, PhD, Postdoc, Faculty).
How we built it
Finding funding opportunities: To find NIH and university-funded projects and fellowships, we built an AI agent to scrape grant websites and extract relevant and actionable details (title, description, eligibility, funding amount, url, deadline, etc.). We also manually input some hard-to-find grants, and grants that are not explicitly listed as funding opportunities.
Semantic search: We use OpenAI's text embedding model to create a semantic embedding of the search term and all funding descriptions. Our model then searches the embedding space of funding opportunities, finding the opportunities whose embeddings most closely match that of the user’s query.
We filter by academic position & group, sort the results by relevance, and display the results.
We used ReactJS, NextJS, Python, and Tailwind CSS for this project, with the support of Perplexity and Windsurf and Perplexity.
Accomplishments that we're proud of
Finding, scraping, and categorizing 1.5k+ grants Building an agent for website scraping across varied websites Building an intuitive front end user interface using NextJS, ReactJS, and Tailwind CSS. Added actionable information and links (i.e., original grants website, add deadline to google calendar, due date countdown) Adding filtering and sorting functionality
What we learned
Semantic search is the best way to find relevant projects to the user’s interests
What's next for Grants Search
We hope to scale Grants Search to other academic institutions across the country and save researchers, student engineers, and faculty thousands of hours every year!
Log in or sign up for Devpost to join the conversation.