Inspiration : Government schemes often exist, but citizens rarely benefit from them because the information is buried in long PDFs or written in complex legal language. Many eligible users don’t even know a scheme exists or how to apply. I wanted to bridge this gap by creating a simple tool that turns policy into actionable guidance.
What I Learned : How to structure semi-structured government data into JSON for AI processing. Using AI (Gemini API) for personalized outputs like match scores and step-by-step checklists. Building a React + Vite app that dynamically renders user inputs and scheme data. Handling AI limitations, like choking on complex or long text, and designing fallback logic.
How I Built It : Collected sample schemes across genres (Education, Health, Agriculture, Employment). Created a JSON database to store scheme details, eligibility, benefits, and required documents. Built a React + Vite frontend with: Input fields for scheme name, user profile, and genre selection. Dynamic rendering of all eligibility criteria. A result section showing AI-generated match score and 3-step checklist.
Integrated Gemini API for personalization: Input: user profile + scheme data Output: { match_score, checklist } Added fallback in case AI fails. Styled the app with dark gradient theme and clear visual hierarchy for readability.
Challenges : Gemini choked on long, dense scheme text, so I had to simplify inputs and chunk content. Rendering dynamic eligibility fields was tricky because each scheme has different criteria. Making the UI responsive and clean while displaying multiple nested fields and AI output.
Future Improvements : Support full-text PDF ingestion with AI summarization. Add scheme recommendation based on location and profession. Multi-language support to reach wider audiences.
Log in or sign up for Devpost to join the conversation.