Inspiration
In rural India, the doctor-to-patient ratio is a staggering 1:811. For millions, language barriers and the lack of immediate triage lead to preventable complications. I created AarogyaVani (meaning "Voice of Health") to ensure that quality medical guidance is never more than a voice command away, regardless of literacy or location.
What it does
AarogyaVani is a multimodal AI health companion that provides instant triage support.
Zero-Click Multilingual Voice: Using Gemini 3’s native audio processing, it automatically detects if a user is speaking Telugu, Hindi, or English and responds in that same language—no buttons required.
Multimodal Triage: Users can import photos of skin rashes, injuries, or prescriptions. The AI analyzes the visual data to provide a [GREEN/YELLOW/RED] urgency alert.
Smart History: It tracks patient chats to identify worsening symptoms over time, providing a longitudinal health record for rural users.
How we built it
The project is built using Gemini 3 Flash for its incredible speed and multimodal reasoning. I developed the frontend using React and Tailwind CSS for a "WhatsApp-style" familiar interface. To handle the scale and offline needs of rural areas, I integrated Firebase for secure chat storage and file hosting.
Challenges we ran into
The Challenge: You likely ran into the "Failed to call the Gemini API" error because Gemini’s safety filters are very strict with medical keywords.
How to say it: "One of our biggest hurdles was navigating Gemini’s strict safety filters. Since we are dealing with health data, we had to refine our prompt engineering to ensure the AI provides helpful guidance without triggering false-positive safety blocks, all while maintaining medical accuracy."
When talking about challenges in a pitch, you don't want to sound like you failed; you want to sound like you are a problem solver. You should frame these as "technical hurdles we overcame" or "valuable pivots."
Here’s how to explain your specific challenges for the ArogyaVani pitch:
- The "Medical Safety" Filter Wall The Challenge: You likely ran into the "Failed to call the Gemini API" error because Gemini’s safety filters are very strict with medical keywords.
How to say it: "One of our biggest hurdles was navigating Gemini’s strict safety filters. Since we are dealing with health data, we had to refine our prompt engineering to ensure the AI provides helpful guidance without triggering false-positive safety blocks, all while maintaining medical accuracy."
- The Language Barrier (Hinglish vs. Native Script) The Challenge: The AI was originally outputting "Hinglish" (Hindi words in English letters) instead of the actual Devanagari or Telugu scripts.
How to say it: "We initially struggled with 'script-switching.' The AI would often respond in Romanized text, which isn't helpful for rural literacy. We solved this by fine-tuning our system instructions to force native script output, ensuring the responses are truly accessible to the local population."
Accomplishments that we're proud of
he project is built using Gemini 3 Flash for its incredible speed and multimodal reasoning. I developed the frontend using React and Tailwind CSS for a "WhatsApp-style" familiar interface. To handle the scale and offline needs of rural areas, I integrated Firebase for secure chat storage and file hosting. Empowering Local Triage: Developing an intelligent [GREEN/YELLOW/RED] alerting system was a major milestone. We successfully tuned Gemini 3 to act as a reliable "first responder," providing clear, decisive action steps that can turn a moment of panic into a path toward recovery.
(still working on this) We take pride in moving from a "sandbox" idea in Google AI Studio to a functional, locally persistent app in VS Code in record time. Overcoming TypeScript hurdles and API integration errors while maintaining an offline-ready architecture proved that advanced AI can be made stable for the unpredictable environments of rural India.
A "Privacy-First" Approach: By utilizing local storage instead of central cloud databases, we’ve built a system where a patient's most sensitive data stays exactly where it should—on their own device.
What we learned
The Power of Multimodal Design: We learned that Gemini 3 Flash is more than just a chatbot; its ability to "reason" across audio and images simultaneously allowed us to build a tool that feels human. We discovered that native audio processing is the key to solving the literacy gap in rural healthcare.
Engineering for the "Dark Zones": This project taught about how gemininano work for offline mode
What's next for ArogyaVani - A Healthcare solution
Integration with India’s Digital Backbone (ABDM) To scale nationally, AarogyaVani must align with the Ayushman Bharat Digital Mission (ABDM).
ABHA ID Integration: Allow users to link their triage history to their unique 14-digit ABHA Health Account. This creates a unified, longitudinal medical record that doctors can access during future visits.
Unified Health Interface (UHI): Integrate with the UHI open network to enable digital consultations and appointment bookings directly from the chat.
- From Triage to "Clinical Agent" Move beyond simple symptom checking to Autonomous Clinical Agents that handle end-to-end workflows.
Predictive Health Analytics: Use Gemini 3 to analyze historical trends (e.g., blood sugar or BP patterns) to predict risks years before symptoms emerge.
Wearable Syncing: Connect the app to low-cost biosensors or smartwatches to monitor vitals like oxygen saturation and heart rate in real-time.
- Strengthening the "Offline-First" Edge To truly serve "dark zones" with no internet, the app will transition to Edge AI.
Gemini Nano Implementation: Deploy core triage and language models locally on the device (using Android's AICore), ensuring 100% uptime without relying on a cloud connection.
Federated Learning: Update the AI model using Federated Learning, where the model learns from local data without raw patient files ever leaving the phone, ensuring maximum privacy.
- Ecosystem Partnerships & Monetization Transition the prototype into a sustainable social enterprise:
Diagnostic Referral Model: Partner with local labs to offer "Home Test" bookings. AarogyaVani can earn a referral fee while users get discounted diagnostic services.
Log in or sign up for Devpost to join the conversation.