VoiceDoc

ABOUT THE PROJECT
SEVERITY OF THE PROBELM RED YELLOW GREEN DEPEDNING UPON THE PROBLEM
Tech Stack of our project
Our Project suggest the path to follow for the problem in local language
Patient Can ask about their problem
Landing Page of our project

Inspiration 700 million Indians live in rural areas where the nearest doctor is hours away. Language is a barrier. Literacy is a barrier. Cost is a barrier. We asked — what if healthcare could speak your language, understand your symptoms, and guide you instantly — no doctor, no internet, no English required? VoiceDoc was born from that question.

What it does VoiceDoc is an AI-powered multilingual voice health triage system built for rural India. Users speak their symptoms in Hindi or 7 regional languages — the system understands, triages severity, and gives instant health guidance. Key features:

Voice Health Triage — speak symptoms, get AI diagnosis guidance MediScan — identify medicines via camera Emergency SOS — one-tap emergency alert 8 Indian Languages — Hindi, Bengali, Tamil, Telugu and more Patient Analytics Dashboard — track health trends over time

How we built it

Frontend: Next.js 14 deployed on Vercel Backend: Flask REST API deployed on Render AI Engine: Groq + LLaMA for ultra-fast multilingual inference Voice Processing: Web Speech API + custom NLP pipeline Database: MongoDB for patient records and analytics Camera/Vision: MediScan module for medicine identification

Challenges we ran into

Python 3.14 and pydantic-core incompatibility on Render — fixed by pinning Python 3.11.9 and pydantic 1.10.2 Hardcoded localhost URLs breaking production — resolved via Vercel environment variables react-leaflet dependency conflict causing Vercel build failures — fixed by cleaning package-lock.json Making voice recognition accurate across Indian accents and dialects with limited training data Balancing AI response speed with accuracy for low-bandwidth rural environments

Accomplishments that we're proud of

Fully functional live product — not just a prototype Real multilingual voice triage working across 8 Indian languages MediScan working in real-time with just a phone camera End-to-end deployment: Vercel + Render + MongoDB in production Built a system that could genuinely save lives in underserved communities

What we learned

Deploying AI systems for Bharat means optimizing for low-end devices and patchy connectivity Dependency management in Python for production is brutally unforgiving — always pin versions Groq's inference speed is a game changer for real-time voice applications Designing for rural users means rethinking every UX assumption — voice-first is not optional, it's essential

What's next for VoiceDoc VoiceDoc is not just a hackathon project — it is a long-term mission to make quality healthcare a right, not a privilege, for every Indian regardless of language, location, or literacy. Phase 1 — Offline First (Next 3 months) Full offline triage capability using on-device compressed LLMs so VoiceDoc works even in zero-connectivity villages. No internet. No problem. Phase 2 — ASHA Worker Toolkit (3-6 months) A dedicated dashboard for India's 1 million ASHA workers — the real frontline of rural healthcare. Automated patient tracking, symptom history, and AI-assisted referral recommendations built into their daily workflow. Phase 3 — Affordable IoT Integration (6-9 months) Sync with low-cost health wearables to capture real-time vitals — heart rate, SpO2, temperature — and feed them directly into the triage engine for more accurate, data-driven diagnosis guidance. Phase 4 — Government Health Stack Integration (9-12 months) Deep integration with Ayushman Bharat Digital Mission and the National Health Authority stack — making VoiceDoc a recognized tool within India's public healthcare infrastructure, not just a standalone app. Phase 5 — All 22 Scheduled Languages (12 months) Expanding beyond 8 languages to cover every scheduled language in the Indian Constitution — ensuring no community is left behind, from Manipuri to Konkani to Dogri. The vision is simple: by 2027, a farmer in rural Jharkhand should be able to describe chest pain in Santali, receive accurate triage guidance in 3 seconds, and get connected to the nearest government hospital — all from a basic Android phone. VoiceDoc makes that possible.