VaxWise

Inspiration

We noticed how often families lose track of vaccinations: a faded paper card, a parent unsure if a booster was ever given, a teenager who was never told they might need one. According to WHO-UNICEF (WUENIC), India still had 6.79 lakh zero-dose children in 2025, down from 9.09 lakh in 2024. Globally the number is 13.5 million. India's U-WIN platform has done impressive work, but it only tracks children up to 16 through government sessions. We wanted to design something that follows the person, not the system, across their whole life and in their own language.

What it does

VaxWise is a proposed multilingual AI vaccination companion.

Smart Life-Course Tracker: builds a personal schedule from date of birth, flagging doses as done, due, or overdue. AI Vaccine Card Scanner: photographs a paper card and uses OCR to convert it into a digital record. "Do I Need This Vaccine?": answers real questions in plain language, grounded in WHO/MoHFW/IAP sources, and always recommends confirming with a doctor. Myth vs Fact: short, sourced cards that address common vaccine misconceptions. Nearest Centre + Reminders: finds the closest free vaccination centre and sends reminders via app, SMS, or voice call. Community Insight Dashboard: an anonymised view of overdue rates and common myths by region. How we approached it

We designed the system as two parts. A rule engine, built on versioned JSON schedules, would decide what is medically due, keeping that logic auditable and separate from the AI. An AI layer would handle OCR, language translation (using AI4Bharat and Bhashini), and a retrieval-grounded assistant that answers questions using verified sources rather than open-ended generation. This is currently our planned architecture, not yet built.

Challenges we anticipate Keeping the AI grounded instead of letting it guess at health answers. Designing for users with low literacy or unreliable internet, not just adding a language toggle. Getting OCR to handle inconsistent, handwritten cards reliably. Positioning VaxWise clearly alongside U-WIN instead of duplicating it. What we learned

Researching this space taught us that the hardest problem isn't technical, it's trust. Any real answer needs a clear source, and "please ask a doctor" has to be a built-in feature, not a gap. We also learned how U-WIN, ABHA, and FHIR fit together, and how to design around that infrastructure instead of ignoring it.

What's next Build a working prototype starting with the scanner and schedule tracker Test OCR accuracy on real paper vaccination cards Expand from 2 to 5 languages with voice support Work toward ABHA/U-WIN interoperability