Inspiration
Our team is entering on the Firecrawl track.
We built a research grant manager in recognition of the constraints on individual grant writers in applying to a wide variety of funding sources. GrantMe is for researchers and labs that find grant-writing to be a key part of their workloads. We recognise that the process of grant writing is often tedious, requiring substantial work to source opportunities, write proposals, and manage their submission through time. Note that this is not just constrained to use in the research grant writing space, but for grant-writing more broadly across the non-profit and start-up sectors. Over 10,000 full-time employees work in this sector in the U.S. alone, costing ~$600M/year to employ.
What it does
Our research grant manager does three things:
- Finds research grants
- Matches available research grants with the goals of the lab
- Writes research grant proposals
In order to do so, it uses a description of the research lab and its areas of exploration (per input from papers) to write and structure research grant applications.
How we built it
GrantMe’s technical framework integrates Firecrawl to dynamically retrieve funding data in markdown from sources like the UKRI funding portal, ensuring real-time updates. Using Camel AI’s multi-agent architecture, it includes specialised agents for grant matching, proposal writing, budget creation, and proposal compilation. Each agent is powered by specific AI models like GPT-4O Mini for OpenAI, orchestrated through Camel’s ModelFactory, allowing precise tasks based on predefined personas. The integration enables parsing of research lab profiles and aligning them with funding requirements. A workforce in Camel coordinates the agents, with task assignments for efficient, end-to-end grant proposal development.
Challenges we ran into
We ran into issues with the Camel-AI codebase, which led to issues with Open AI’s API, even though it wasn’t being used. The hackathon session lasted only two hours, which means many improvements remain just ideas.
Accomplishments that we're proud of
We are proud of having developed a working architecture for the development of one-page submission-quality research grant proposals.
What we learned
We learned a substantial amount about CamelAI.
What's next for GrantMe
Moving forward, we want to expand the project to multiple pages/complete research grant proposals, and streamline the submission process also.
Built With
- camel
- firecrawl
- python
Log in or sign up for Devpost to join the conversation.