AWS Public Sector Blog
AWS and UNC researcher build a prototype agentic AI tool to streamline grant funding
Federal funding cuts and fierce competition are forcing US researchers to look beyond their usual grant sources. Dr. Adam Kiefer, a researcher at the University of North Carolina at Chapel Hill (UNC), knows this challenge well. He regularly spent hours reformatting and reworking similarly themed grant proposals for different agencies, turning an eight-page National Institutes of Health (NIH) proposal into a one-pager for the Department of Defense, then restructuring it again for the Defense Advanced Research Projects Agency (DARPA). Meanwhile, grant opportunities he’d never heard of were expiring in databases he didn’t know existed.
Kiefer isn’t alone. Researchers spend 44% of their time on administrative tasks rather than on actual research, with proposal preparation taking the longest. Finding new grants and creating proposals for them means manually checking Grants.gov, NIH, National Science Foundation (NSF), and dozens of other sources. Early-career researchers face even steeper barriers without established networks or agency listservs.
To address this, Amazon Web Services (AWS) worked with Kiefer as a subject-matter expert to develop the Grant Research Opportunity Wizard (GROW). Available on the AWS Samples GitHub repository, this open-source generative AI tool helps researchers develop proposals faster—and is now evolving to include agentic capabilities that will automatically discover relevant grant opportunities worldwide.
How the idea for GROW emerged
The concept of GROW grew out of years of collaboration between the AWS Academic Research team and Kiefer on cloud-native research capabilities. During these conversations, Kiefer kept mentioning the same frustrations: the endless cycle of reformatting grant proposals and the difficulty of finding opportunities beyond familiar funders. Sometimes those frustrations cut their meetings short or canceled them entirely because Kiefer had a deadline to meet.
“It just got so burdensome,” Kiefer said. “I felt like every day I was getting asked to put something new into my proposal—another section, another format. I realized there’s got to be a more efficient way to do this.”
The AWS team recognized these weren’t just Kiefer’s problems—they represented a gap that technology could address.
That realization sparked an idea. Kiefer worked side by side with the Academic Research business development and solutions architect team to build the first version of GROW, focusing first on proposal development. “They were coming back with ideas that were well beyond even what I had been considering. It was exciting,” Kiefer said.
Streamlining NIH proposals
The result was GROW version one, released as open-source on the AWS Samples GitHub repository in late 2024. It helped researchers draft NIH proposals more efficiently using generative AI, ingesting a researcher’s publications and draft materials, then generating proposal content structured to NIH requirements.
“As a researcher, it’s going to take me a minimum of three weeks to think through a project concept and get something down on paper,” explained Jack Fenwick, AWS business development leader. “If I can start with a first draft that matches my expertise against solicitation requirements, instead of a blank page, I can save weeks building out a full proposal.”
Multiple research organizations across the country, including major universities, deployed GROW version one as a proof of concept. Their feedback revealed a critical gap: With intensifying competition, researchers needed help not just with writing proposals but also with finding new grant opportunities beyond their usual sources.
From proposal drafting to autonomous discovery
This feedback drove a major shift. GROW version two, moves from generative AI that responds to prompts to agentic AI that autonomously matches global funding opportunities to a researcher’s expertise. Powered by Amazon Bedrock, Amazon Bedrock AgentCore, and Amazon OpenSearch Service, the tool also includes built-in guardrails to protect sensitive research data, and organizations retain full control of research profiles and proposal content.
Once installed in a research environment, GROW version two, uses AI agents to autonomously crawl funding databases multiple times per day, including Grants.gov, federal agency websites, and international sources like Horizon Europe. Researchers can build profiles by uploading a CV or linking existing online profiles, then create multiple profiles for different research lines. They can also customize their search by specifying expertise, career stage, target topics, and preferred funding agencies.
Kiefer anticipates that flexibility will be key for his research. “I’ll make sure my profile depicts what I think I should highlight for this particular research line and then let the agentic AI go to work for me,” he said.
Using knowledge bases, GROW matches opportunities to researcher profiles, publications, and expertise, surfacing grants that typical searches would miss and providing them with a matching score to identify funding opportunities that most closely match their profile. Once a researcher discovers the right opportunity, GROW does much of the legwork: it pulls best practices and guidelines for each funder and generates proposal templates formatted to their specific requirements. The result: a head start on both discovery and proposal development.
Finding the right grant opportunities faster
For Kiefer, GROW opens possibilities he couldn’t access before. “I’ve got collaborators in Europe. I don’t know anything about the European system,” he said. “GROW can go in there and find those opportunities—I’m not even sure I would know where to begin if I had to do that manually.”
Beyond geographic reach, GROW surfaces funding opportunities in areas researchers might not typically explore. Kiefer already knew his eye-tracking research on basketball performance could apply to military contexts like soldier battlefield awareness, but he wasn’t familiar with military funding mechanisms. “GROW connects me to those opportunities,” he said. “Without GROW, I’d be spending hours trying to uncover these on my own, and I would still probably miss some.”
But expanded discovery doesn’t mean chasing every opportunity. GROW helps researchers focus strategically. “What GROW does is give you the precision,” Kiefer explained. “You’re casting that wide search net, but then it’s bringing in the high probability matches so that you can focus your attention on the ones that are going to give you the best chance.”
That focus translates to smarter effort, not more effort. “I don’t want to write more grants,” Kiefer added. “I want to dedicate more time to writing proposals where I have a much higher probability of success, versus wasting it going after things that maybe I just stumbled upon.”
For Fenwick, this represents a fundamental shift in perspective: “It grows the aperture significantly from one that is a narrow view, because of personal experience, to a much wider perspective on what is out there, what is possible.” Early estimates suggest GROW could help researchers become up to 30% more efficient, freeing up time to reinvest in refining their grant approach and developing more strategic, competitive submissions.
Open source for any research organization
That efficiency gain will soon be available to the broader research community. GROW version two is targeted for release as open-source on the AWS Samples GitHub repository in early 2026. By releasing GROW as an open-source prototype, AWS aims to help research organizations of all sizes access the same discovery and proposal tools—enabling researchers to focus on their research and institutions to grow research output across the enterprise. Universities, government labs, nonprofits, and research institutes can deploy, customize, and extend their capabilities.
These changes are just the beginning. GROW’s roadmap now extends beyond grant discovery to proposal submission, budget development, and post-award management.
Ultimately, the collaboration reflects what’s possible when researchers and technologists tackle shared challenges together. “This is just one example of what can happen when we’ve got an academic researcher working with an industry powerhouse like AWS,” Kiefer said. “It can help us change the way we think about a lot of these manual, time-consuming inefficiencies in the research system and overcome them for the broader community.”
Contact AWS today to learn more about how we support higher education and research.