Sahayak AI

Inspiration ## Inspiration

Meet Ramu bhaiya — 52 years old, small farmer in UP, 2 acres of land. He is eligible for ₹6,000/year from PM Kisan Samman Nidhi. He has never received a single rupee. He doesn't know the scheme exists.

This is not a rare story. This is India's daily reality for 8 crore farmers and 60 crore rural citizens who are legally entitled to welfare but have never accessed it.

India has 400+ government welfare schemes worth ₹15 lakh crore. Yet ₹9 lakh crore goes unclaimed every year — not because the money isn't there, but because nobody told Ramu bhaiya it was.

What it does

Sahayak AI is a free, voice-first, Hindi AI assistant that helps any Indian citizen — regardless of literacy, language, or tech skills — discover, understand, and apply for the government welfare they deserve.

Key features:

🎤 Voice input in Hindi — Press a button and speak. No typing, no English required
📋 Scheme Finder — Search 6+ major schemes with full eligibility, documents, and apply steps
🤖 AI Chat — Ask anything in Hindi, get instant structured answers
🛡️ Scam Guard — Real-time fraud detection, alerts users before they share OTPs
🧮 Benefit Calculator — Converts yearly benefits to monthly so users grasp real value
📊 Application Tracker — Track status of all applications
🔊 Text-to-Speech — Every answer read aloud in Hindi for low-literacy users

How we built it

Frontend: React 19 + Vite 8 + Tailwind CSS v4 + Framer Motion
Backend: Node.js + Express (lightweight proxy server)
AI: Google Gemini API with smart offline fallback for rural low-connectivity areas
Voice: Web Speech API (Hindi recognition hi-IN + synthesis)
Design: Mobile-first, warm saffron palette, Baloo 2 + Mukta fonts (Indian typography)

The entire stack runs on free tiers. This app costs ₹0 to operate.

Challenges we ran into

The hardest part was designing for users who have never used a smartphone app before. Every decision — large mic button, Hindi-first text, no jargon, voice-primary interaction — was made for a 60-year-old rural user, not a tech-savvy one.

API reliability was another challenge. We built a smart offline fallback system so the app answers questions about all major schemes even without internet — critical for rural India.

Accomplishments that we're proud of

Built a fully working 6-feature app in one hackathon
AI correctly identifies 8 types of welfare needs and scam patterns
Scam detection catches real fraud message patterns used in the wild
Voice input works in Hindi on mobile browsers without any third-party SDK

What we learned

Government scheme data is shockingly hard to structure. Every scheme has different eligibility rules, document requirements, and processes. Building a unified, user-friendly layer on top of this complexity was the core product challenge.

We also learned that the real barrier to welfare access is not awareness alone — it's a combination of language, literacy, trust, and process complexity that needs to be solved together.