JolSetu

• 

  JolSetu home - the arsenic crisis in numbers, and the 3-step bridge to safe water (Bangla & English, offline-first)

• 

  Live map - 6,593 real Araihazar tubewells; the model flags where to test next; offline walking routes to verified safe wells

• 

  AI kit reader - photograph the $1 Gutzeit kit; explainable color-match verdict in seconds, no more chart-guessing

• 

  Fraud-proof check-in - QR-stamped, GPS-geofenced to the well; simulated cheater rejection keeps the map honest

• 

  Predict & Prioritize - where the map and model disagree, communities test next; model card is public

• 

  Bangla UI - spoken (TTS) results and full Bangla interface for low-literacy and offline rural users

💧 JolSetu — The Water Bridge

AI reads a 20-taka arsenic test kit. Photograph the Gutzeit reaction pad next to a white reference card; on-device color science (white-reference normalization + CIELAB matching) reads the verdict and speaks it in Bangla.

• 🌐 Try it: https://rudra496.github.io/jolsetu
• 💻 Source: https://github.com/rudra496/jolsetu
• 📚 Evidence: every claim mapped to a verified source in the repo's EVIDENCE.md

The problem

• 20 million Bangladeshis drink water above the national arsenic standard; WHO called arsenic "the largest poisoning of a population in history"
• Testing is once-in-decades, readings are human color-chart guesses, and when a well fails, families have no way to find the nearest verified safe one

The four pillars

1. Test Kit AI — colorimetric reading validated against 943 real kit-vs-lab comparisons; attacks the 37% false-condemnation rate of human chart reading
2. QR-stamped live map — every check-in is HMAC-signed, GPS-geofenced and duplicate-checked; trust is a design constraint
3. Predict & prioritize — a logistic risk model (AUC 0.67, disclosed) flags where model and map disagree; the crowd tests there next
4. Safe-well routing — walking directions to the nearest verified safe wells, offline via service worker

Built on real data

All 6,593 prototype wells are real measurements from the Araihazar upazila survey via CUAHSI HydroShare (DOI-linked).

Recognition

Completed and published for the Xylem Global Student Innovation Challenge 2026; entered into the IEEE OneAquaHealth challenge as a team build.

Tech stack

JavaScript · on-device computer vision · PWA/service worker · HMAC-signed check-ins · logistic risk model

Built With

• computer-vision
• javascript
• machine-learning
• pwa