Bridging India’s Welfare Gap: Integrating RAG, Eligibility Matching and Multimodal Verification in One Civic AI Platform
Bridging India’s Welfare Gap: Integrating RAG, Eligibility Matching and Multimodal Verification in One Civic AI Platform
Authors:
Abhay Sharma, Om Wallalwar , Prof. Avinash Taskar
Sandip University, SOCSE, B.Tech CSE
[Nashik, India]
Abstract - This paper presents SchemesSense, an AI-powered web application designed to help Indian citizens discover government welfare schemes through intelligent conversational assistance. India runs over 300 central welfare programs — health, education, housing, agriculture — yet a large share of eligible citizens never see any benefit. The reasons are not complicated: English-only content, confusing eligibility rules, documents that nobody explains, and government portals that assume a level of digital comfort most users do not have.SchemesSense tackles this with five modules built into one platform. A document verification engine uses Google Gemini 2.5 Flash to visually inspect uploaded government documents — Aadhaar, PAN, Income Certificates — and return a verdict with confidence scores and tampering signals. A RAG-powered chatbot retrieves scheme data before generating any response, keeping answers factual. A smart eligibility checker takes user details and returns a ranked list of schemes they actually qualify for. A 30-second video explainer covers each scheme in plain language for users who will not read dense text. A community blog lets citizens who have applied share what the process actually looked like.The backend runs on Spring Boot 4 with JWT authentication, PostgreSQL for user data, and Firebase Firestore for scheme content. The frontend is React 19 with Vite 7. Everything is deployed — frontend on Vercel, backend in Docker. Testing across 50 queries showed 90% accuracy in scheme recommendations, with document verdicts returning in under two seconds. Hindi and English are both supported throughout
Key Words : civic technology, welfare scheme discovery, retrieval-augmented generation, document verification, eligibility matching, conversational AI, multilingual interface, Indian e-governance, Gemini API, community knowledge sharing.