AI सखी

For healthy, thriving women...

According to the McKinsey Health Institute, while there has been significant rise in life expectancy for both men and women over the past 2 centuries, women spend more of their lives in poor health and with degrees of disability (the “health span” rather than the “life span”). This affects her ability to be present and/or productive at home, in the workforce, and in the community and reduces her earning potential ("Closing the women’s health gap: A $1 trillion opportunity to improve lives and economies").

While the report provides comprehensive data on healthcare gap for women, and makes an economic argument for addressing the data gaps, investment and innovation, Dr. Elinor Cleghorn points out in her powerful book "Unwell Women" how health science itself has failed women over centuries through misunderstansing and misdiagnosis of women's health issues. If we combine this with the lack of healthcare access for women underserved by the healthcare system, the urgency of action becomes even more evident.

Technology is seen as a great equalizer in settings of poor infrastructure - be it in healthcare, education or the economy - by acting as a powerful distribution channel. More women globally have their own phones (even in rural areas) - now more than ever before. My project builds upon this key insight and brings the power of artificial intelligence to this problem.

“AI सखी” (the word सखी "sakhi" means friend in Hindi) is a generative AI-powered expert system that can be delivered to under-served communities digitally. It can be used directly by women, or by rural healthcare workers in villages.

The system is based on Large Language AI Models (LLMs) from OpenAI. It augments generalist knowledge of these models with domain-specific knowledge bases and can accommodate targeted, population-specific data.

The system can operate in regional languages (such as Hindi, spoken by a large number of villagers in India) and has text-to-speech / speech-to-text capabilties.

The system also acts as a platform for longitudinal data collection at the point of care - both for convenient physician access when intervention is needed - and for developing statistical and predictive tools for women’s health.

Overview and First Prototype

The following video provides an overview of the project, and showcases the very first prototype demonstration of the system.

Here are some pictures from my very first high school presentation :-)

The AI Sakhi Stall !

With school superintendent Dr. Wenzel

Collaborating for mutually reinforcing missions

I've had the privilege of collaborating with the Pardada Pardadi Education Society and their Prana Clinics in India, and Ms. Miki Sunguza's A La Carte Health in the US, to pilot my work. These organizations are making profound contributions to girls' and women's health around the world. Our pilots have focused on AI-powered clinical assistance, and providing health recommendations to women - grounded in trusted medical knowledge bases and expert guidance on women's health. The prototypes shared below provide insight into this work and the pilots.

Prototype 2 using Knowledge Graphs for advanced reasoning

In version 2 of the system, knowledge graphs were added to anchor the AI model's reasoning - grounding responses in authoritative domain knowledge. The system uses NHS data to create the knowledge graph.

Prototype 3 for clinical assistance in Prana Clinics

In this version, the system was enhanced to act as a clinical assistant for doctors in Prana clinics. Using voice recognition and a generative AI model the system is able to transcribe complex domain information and capture it in the clinic's backend systems (eg. Salesforce) in the right structure and fields, significantly reducing the effort from doctors and nurses. This can be used as an early entry point prior to advanced diagnostic assistance.

Page updated

Google Sites

Report abuse