Sanjeevani AI+

An offline-first AI assistant for rural healthcare—matching blood donors, detecting symptoms, and guiding care where it's needed most.

Comment

Inspiration

Thalassemia patients in rural India face a dire lack of timely blood transfusions due to slow donor matching, limited access, and minimal tech support for ASHA workers. We were inspired by the mission of Blood Warriors Foundation and driven to build a scalable, AI-powered solution that empowers frontline healthcare workers—even without internet. Our goal: deliver timely care, bridge systemic gaps, and save lives.

What it does

Sanjeevani AI+ is an offline-first, multilingual AI assistant that:

  • Instantly matches blood donors based on location and type
  • Triages symptoms via bilingual voice/text chatbot (Hindi + English)
  • Detects skin complications using a lightweight CNN model
  • Works fully offline and syncs when back online
  • Enables ASHA workers and rural users to access critical care services with minimal training

How we will build it

We developed a cross-platform mobile app using:

  • React Native + Expo for fast UI prototyping
  • TensorFlow Lite for on-device symptom triage and skin condition detection
  • DistilBERT and rule-based NLP for triage classification
  • Whisper for accurate multilingual voice input
  • Firebase for lightweight real-time data sync and user auth
  • OpenStreetMap + rule-based logic for donor geo-matching
    We prioritized modularity and offline performance to ensure real-world applicability.

Challenges we ran into

  • Optimizing AI models for offline, low-resource devices without compromising accuracy
  • Ensuring reliable voice input across different accents and dialects
  • Building an intuitive UX suitable for semi-literate users
  • Integrating multiple services (triage, scan, matchmaking) into one cohesive, lightweight app
  • Balancing MVP scope with long-term scalability

Accomplishments that we're proud of

  • Built a fully functional offline AI healthcare assistant in just 2 days
  • Seamlessly combined voice-based NLP, computer vision, and geolocation in a single platform
  • Designed the app around real-world usability by ASHA workers and rural patients
  • Created a scalable foundation that can plug into national systems like e-RaktKosh

What we learned

  • Lightweight AI (like TFLite and DistilBERT) is highly effective when tailored for edge environments
  • Offline-first design isn’t just a technical choice—it’s a human-centered necessity
  • Simplicity and empathy in UI/UX go a long way when serving low-literacy user bases
  • Voice input, when done right, can break barriers in rural healthcare access

What's next for Sanjeevani AI+

  • Expand language support to more regional dialects
  • Integrate with real-time blood bank systems (e.g., e-RaktKosh, hospital APIs)
  • Add features like vitals tracking, appointment reminders, and caregiver coordination
  • Pilot deployment with a healthcare NGO to gather real-world feedback
  • Open-source the core platform to drive grassroots adoption and co-development

Built With

Share this project:

Updates