This is a submission for the Hacktoberfest Open-Source AI Challenge Week 2: Side Hustle.
you
What I Built
Finding funding for something you're passionate about should be exciting. Whether it's a research grant, a conference travel award, or a fellowship that could open the door to your next opportunity, financial support can make an enormous difference. Unfortunately, finding an opportunity is often easier than understanding whether you actually qualify for it.
As a researcher myself, I've experienced how frustrating this process can be. You might discover an award that seems perfect, only to spend hours navigating eligibility requirements, exceptions, and application instructions before realizing that one small condition could change everything. For students, independent researchers, and people without access to dedicated funding support, this process can be especially overwhelming.
That's why I built FundPilot AI, an open source funding assistant powered by Google's Gemma that helps people understand funding requirements and prepare for applications with greater clarity.
I wanted to create something I would genuinely use myself. With FundPilot, users can paste the public requirements from a funding announcement, answer a few straightforward questions, and receive an organized overview of what appears to match their circumstances, what might be a problem, and what still needs clarification.
One of my favorite features is the Application Readiness Dossier. Rather than simply identifying something as uncertain or potentially conflicting, FundPilot helps users understand what to do next. The downloadable dossier brings together possible obstacles, application materials, important dates, supporting evidence, and suggested questions to ask the funding organization.
For example, if a conference travel award requires applicants to be the first author of a presentation, FundPilot can recognize that requirement, compare it with the information provided, and highlight a potential conflict. Instead of leaving someone wondering what that means, it connects the finding to the original announcement and suggests what should be verified before applying.
Accessibility has been central to how I've approached this project. I wanted the experience to feel approachable even for someone unfamiliar with programming or AI. The latest version features a cleaner results dashboard, questions suggested from the funding text, and expandable sections that make the information easier to explore without becoming overwhelming.
FundPilot does not make official eligibility decisions. My goal is to help people understand opportunities, recognize what needs attention, and feel more prepared to take their next step.
I also see an exciting possibility for this to become an affordable funding preparation service, particularly for individuals and smaller organizations without grants offices. My initial business idea is $9 per readiness report or $39 per month for a funding workspace. These are proposed prices, not existing paid services. I want to validate whether FundPilot saves people time and provides enough value before introducing payments.
Demo
This 96 second demonstration shows FundPilot running with Gemma and turning public funding requirements into practical next steps. It walks through matching requirements, potential conflicts, missing information, and the Application Readiness Dossier.
The recording includes four genuine Gemma analyses and eight verified report downloads. Applicant information is fictional, and the selected public passages are examples rather than complete reviews of the funding programs.
You can also explore an example Application Readiness Dossier or visit the latest release.
Code
FundPilot is built using Python, Gradio, Ollama, and Google's Gemma 3. The application code is publicly available under the MIT license, while Gemma's model weights are downloaded separately under Google's terms.
The repository includes the working application, tests, example reports, and documentation explaining how the system was developed and evaluated.
FundPilot currently runs locally, so users need to install the application and Ollama. Making setup simpler and eventually offering a browser based experience are important future goals, especially because I want the tool to become accessible to people without technical backgrounds.
How I Built It
I started with a simple question: could a small AI model help someone understand complicated funding requirements while still allowing them to verify the original information?
I used Gemma 3 1B, running locally through Ollama, to help identify requirements, exclusions, deadlines, and other important details in public funding announcements. Python then compares supported conditions with the information supplied by the user and organizes the results into an understandable report.
One decision I made early was to preserve the original evidence. When FundPilot highlights a requirement, the supporting passage comes directly from the announcement rather than being written by the AI. This matters because an incorrect interpretation could influence whether someone decides to pursue an opportunity.
During development, I discovered that understanding funding language is much more complicated than simply summarizing a document. Some announcements contain exceptions, alternative ways to qualify, or conditions that depend on information the applicant hasn't provided. Instead of hiding that uncertainty, I designed FundPilot to make it visible and suggest what needs further clarification.
The application also includes a separate Rule baseline (NO AI) option, allowing users to analyze the same text without relying on Gemma. My evaluations against selected public funding passages showed that Gemma still has room to improve and does not consistently outperform the simpler rules. The evaluation references also need independent human review. I experimented with a larger Gemma model, but its slower processing and mixed results didn't justify replacing the smaller version.
As I continued developing FundPilot, I realized that identifying a requirement was only part of what users needed. They also needed help understanding what the information meant for them and what action to take afterward. That realization inspired the verification system, the clearer results overview, and the Application Readiness Dossier.
The latest version has 147 passing automated tests, along with successful checks of navigation, downloads, input handling, and layouts on smaller screens. These checks establish important aspects of software reliability, although they don't guarantee that every AI interpretation is correct.
For anyone interested in the technical details, I've included the evaluation results and limitations in the repository.
Why Does Open Innovation Matter?
One of the things I love most about open source technology is that it allows people to build solutions around problems they experience firsthand.
I didn't start FundPilot simply because I wanted to build another AI application. I started it because I know how much funding opportunities can matter and how discouraging it can feel when understanding those opportunities becomes another obstacle.
Using Gemma locally gave me the freedom to experiment, understand where the model made mistakes, and continue improving the application without paying for every analysis through a commercial AI service. Public announcement text is processed locally, while the applicant's answers remain within the Python application. This gives me more control over the process, although it doesn't mean the model is automatically more accurate or that running it has no hardware costs.
Making FundPilot open source also allows others to examine the approach, learn from its development, and potentially contribute improvements. I hope it can eventually become a useful resource for researchers, students, nonprofits, and other people navigating funding applications without extensive institutional support.
As someone whose own research journey has depended on grants, fellowships, and conference opportunities, this project feels especially personal. I know how exciting it can be to find an opportunity that could move your work forward, and I want more people to experience that excitement without feeling overwhelmed by the application process.
I don't believe AI needs to provide every answer to be valuable. Sometimes, helping someone identify an important requirement or understand the right question to ask is just as meaningful.
My hope is that FundPilot can make funding information easier to understand, preparation less intimidating, and opportunities more accessible to the people pursuing them.
My First User
One of the most encouraging parts of developing FundPilot was receiving early feedback from someone outside the development process. They appreciated the clean visual design and recognized the verification system as one of the project's strongest features.
Their observations helped guide my latest improvements. I focused on making the experience easier to navigate through a simpler questionnaire, a clearer results dashboard, and supporting evidence that users can explore when they need more detail.
This was an early qualitative review rather than a completed independent usability study. The participant did not pay for the product, and I have not yet measured time savings. I'm excited to continue gathering feedback and learning whether FundPilot helps people approach funding opportunities more efficiently.
Prize Categories
Best Use of Gemma
FundPilot uses Google's Gemma 3 to help interpret public funding requirements and turn complicated information into a clearer, more accessible application preparation experience!
Top comments (0)