Swasthya Sethu

SwasthyaSetu: Portable, Predictive Care for Migrant Workers

What inspired us

We were inspired by a simple but painful reality: migrant workers move often, but their health records do not. Paper prescriptions get lost, language changes across states, and follow-ups break when people relocate. We wanted to build a system where care continuity is not tied to one clinic, one language, or one city.

How we built the project

We designed SwasthyaSetu as a connected platform with six working modules:

1. SwasthyaID (QR Health Passport)
   A QR-based digital identity that lets any authorized provider access a worker’s medical timeline with consent.

2. BhashaSehat (Voice + Language AI)
   Symptom capture through voice, with multilingual speech-to-text and translation so patients can speak naturally.

3. Document AI (OCR + Structuring)
   Prescription/report scanning that extracts key fields and converts paper documents into searchable digital records.

4. KaamSuraksha (Occupational Risk AI)
   Risk scoring for work-related illnesses based on occupation, exposure profile, and medical indicators.

5. SehatSetu (Care Continuity Engine)
   Health tracking from source to destination location, with follow-up state syncing.

6. Notification System
   Medication reminders and follow-up alerts via SMS/WhatsApp/IVR for better adherence.

For occupational risk scoring, we used a simple interpretable model:

[ R = \sigma\left(w_1E + w_2D + w_3P + w_4H + b\right) ]

where (E)=exposure intensity, (D)=duration of exposure, (P)=PPE non-compliance factor, (H)=health history indicator, and (\sigma) is the sigmoid function.

What we learned

• Healthcare technology must be human-first, not feature-first.
  
• Language accessibility is as important as clinical accuracy.
  
• Consent and privacy design should be built from day one, not added later.
  
• A small, reliable workflow beats a large but fragile prototype in hackathons.

Challenges we faced

• Noisy voice input and dialect variation reduced speech accuracy.
  
• Prescription OCR quality varied across handwriting styles and photo quality.
  
• Cross-location continuity needed strong identity and sync logic.
  
• Scope pressure forced us to prioritize an MVP over many ambitious features.

How we handled those challenges

• Added fallback flows for manual correction after voice/OCR extraction.
  
• Used structured templates to normalize inconsistent medical documents.
  
• Built a clear consent checkpoint before sharing records across providers.
  
• Focused demo scope on core journey: register, capture, digitize, risk-score, follow-up.

Outcome

SwasthyaSetu demonstrates that portable identity, multilingual AI, document digitization, and occupational risk intelligence can work together to deliver continuous care for migrant workers across states.

Built With

• abha-compatible-identity-linking-(prototype-flow)-dev-tools:-git/github
• cloud-object-storage-for-document-images-apis/integrations:-whatsapp/sms-notifications-(twilio)
• languages:-python
• occupational-risk-scoring-model-database:-postgresql-cloud/platform:-databricks-free-edition-(serverless-notebooks-+-sql-warehouse)
• ocr-pipeline
• postman
• qr-code-generation/scanning
• rest-apis-ai/ml:-speech-to-text
• sql-frontend:-react-(vite)
• tailwind-css-backend:-fastapi-(python)
• text-to-speech
• translation
• typescript
• vs