Digital Health

My work on digital healthcare in India and Bangladesh is focused on advancing health literacy for the global majority, and has led to our tech venture incubated at MIT, and now with UNICEF.

Sakhi is my team’s AI-powered digital health literacy platform that revolutionizes preventative maternal care for women in rural and urban India. Our solution addresses the critical issue of maternal mortality (103 deaths per 100,000 live births in India) by targeting one of the systemic drivers of this challenge: the lack of health literacy.

Sakhi offers the following features:

Multilingual WhatsApp-based Conversations

Sakhi is powered by locally adapted LLMs designed to operate in the dialects that its users are most familiar with (e.g. Marathi, Hindi, Bangla). Unlike traditional chatbots that lightly wrap LLM API calls, Sakhi is built with user-informed product guardrails operating over human-curated knowledge bases. It provides personalized health information and ANC reminders to expectant mothers without suffering from hallucinations.

Behavioral Analytics Reporting

Sakhi provides a real-time analytics dashboard for NGOs to track behavioral changes and monitor program impact includes prioritizing at-risk mothers based on their input queries and requests, identifying emergent health-related issues across a region, and compliant health information systems data.

Integrated into Local Health Infrastructure

India has a robust network of AAA workers Sakhi offers actionable AI-driven insights for these frontline AAA workers to improve program delivery. We open direct communication channels between ASHAs and the expectant mothers using our platform to facilitate structured and moderated health-seeking conversations while bridging the physical divide that restricts access to healthcare services.

Check out our presentation and platform here!

Pilots Ongoing

• During our participation in MIT’s flagship social innovation accelerator, Delta V, we conducted an 8-week pilot with an Indian nonprofit organization serving tens of thousands of people across hundreds of villages, running programs for maternal care, HIV-positive individuals, LGBTQIA+ communities, domestic violence survivors, and more.
• During this pilot, we interviewed 35 healthcare workers, ASHA workers, and Aadhar Sanstha volunteers to understand their roles and identify how Sakhi could assist them in providing preventive care services to pregnant women.
• We developed Hindi and Marathi conversational capabilities within the platform, and with information vetted by practicing medical professionals, we are launching pilots to serve the local population in Maharashtra, India.

Human-centered Design for Conversational AI Systems

We also spoke with women in the region to learn about their needs, daily lives, and how Sakhi could effectively serve as an intervention. A Minimal Viable Product (MVP) of Sakhi was tested by 29 pregnant women over 15 days, and we have created a report including the user engagement metrics gathered during this pilot. Presented below is an exploration of the data from a sample of early users of the product that is intended to provide an empirical discussion of how we track behaviors with our system.

• Note that this is anonymized data obtained with user consent.

Menstrual Health Offshoot Pilot

• We recognized that our platform could be utilised beyond maternal health, if structured appropriately and adapted to a new use-case.
• Thereafter, we adapted it to include conversational capabilities in Bangla, and are launching a local pilot with indigenous communities in Bangladesh in order to advance menstrual health literacy.