1 of 8

Sakhi: Automated Menstrual Health Worker

Team SimPPL and AI4Good

sakhi.simppl.org

2 of 8

Menstrual Health Information Is Not Widespread

Across Bangladesh

14 million adolescent females:

  • 50% unaware of cycles at menarche
  • 14% suffer from reproductive tract infections
  • 40% miss school each month
  • Missing on average 2.5 schooldays monthly

3 of 8

A Solvable Problem – examples of early success!

But they are extremely cost and labor intensive, not scalable!

Can technology help?

WASH in-school interventions have led to:

  • 32% higher likelihood of girls maintaining proper menstrual hygiene
  • Drop in school absentee rate from 2.5 days to 1.5 days per month

4 of 8

What about a Chatbot?

  • Most don’t support Bengali or don’t perform well
  • Even OpenAI admits ChatGPT frequently hallucinates information!
  • Most users are unfamiliar with a new chatbot UI, website, login, etc.
    • Huge learning curve limits adoption
  • Expensive and technically challenging to develop and deploy new model for a niche use case!

5 of 8

AI4Good’s WhatsApp Menstrual Health Chatbot

  • Chats in Bengali, “Beng-lish”, and English
  • Disseminates only local, verified and factual data
  • Scalable to over 44 million users on Whatsapp!
  • Developed in collaboration with partners in Bangladesh
  • Tech team SimPPL: Females + Bengali-speaking students who are Indian NLP experts winning grants from Google, Mozilla
    • Successfully Beta Tested by Bangladeshi Healthcare Professionals
    • Building an automated test suite + audio conversations!

(click here for video)

6 of 8

Impact and Next Steps

Launching to alpha users in a pilot experiment!

  • Expression of Interest by Meta and WhatsApp (Public Health)
  • Featured on the Google Cloud Blog (upcoming)
    • SimPPL, led by Swapneel, won a Google Cloud Research Award
  • Research Support confirmed: MIT and Boston University
  • Advisor confirmed: OpenAI’s fmr. Head of Trust and Safety

7 of 8

Deployment (ongoing)

IRB-backed research study in Bengali women and girls aged 10-20 years of age to determine utility provided through this service.

​

Analyze the educational impact through survey-based research.

​

Develop a qualitative and quantitative report for the questions answered and unanswered by the scope of the chatbot based on materials provided.

​

Prepare for adapting the chatbot to broader scope in public health and multiple languages in India and Bangladesh.

8 of 8

Special Thanks

Arpa Paul, Bengali Healthcare Professional

​

Pratyay Banerjee, ML Translation Engineer, West Bengal (India)

Raghav Jain, Research Lead, India

Mrunmayi Parker, WhatsApp Dev Lead, India

Nahush Patil, WhatsApp Dev Lead, India

Augustina Abisola, Period Activist , Nottingham, UK

Chahrazad Belheine, Period Activist, Paris, France

Armir (Lead), Sadhli (Ops), Swapneel (Tech)