Jacaranda Health Scales Equitable Maternal and Newborn Health Using AWS

Founded in 2016, Jacaranda is a Kenya-based healthcare organization that is dedicated to helping mothers experience safe and dignified pregnancies and childbirths. Jacaranda partners with governments and public health systems across multiple African countries—including Kenya, Ghana, and Nigeria—to deploy scalable, low-cost technology solutions that improve quality of care and outcomes for mothers and infants.

One of Jacaranda’s most impactful projects is an SMS-based service called PROMPTS, which supplies clinically accurate health information to pregnant mothers and new parents and answers their questions in near real time. Jacaranda enhanced and scaled the solution with support from the AWS Health Equity Initiative. At first, Jacaranda set out to build a one-way push notification service to send relevant health information and reminders to new and expectant mothers who might not have internet access. However, expectant parents quickly began responding to the texts with questions about their pregnancies.

In response, Jacaranda created a help desk that was staffed by a single agent who prioritized and answered up to 100 questions each day. “Many people in rural Kenya might live more than 1 hour away from the nearest health facilities,” says Jay Patel, director of technology at Jacaranda. “With no access to the internet, they lack ways to get information about issues they experience during pregnancy. PROMPTS is a lifeline that they can reach out to and get reliable, helpful information from.”

As more people signed up for PROMPTS, question volumes rapidly increased. Jacaranda expanded its help desk, but agents could no longer manually sort and answer questions. In response to this growing demand, Jacaranda implemented an artificial intelligence (AI) model to prioritize incoming messages and respond to nonurgent questions. “AI helped us shift human intelligence to the urgent clinical questions,” says Patel. “So, we can quickly direct parents with critical needs to the right care.”

Jacaranda built its messaging service and data pipelines using HAQM Aurora, a relational database management system built for the cloud with full MySQL and PostgreSQL compatibility, and HAQM Redshift, which uses SQL to analyze structured and semistructured data across data warehouses, operational databases, and data lakes. This infrastructure worked seamlessly with Meta Llama 3, an open-source AI model.

After teaching the model to understand Swahili, Jacaranda trained it using a database of more than 1 million health questions and answers. Now, the model can respond to 70 percent of the questions that PROMPTS receives, such as low-priority inquiries around diet and exercise. “AI doesn’t get tired, frustrated, or bored when doing repetitive tasks,” says Patel. “It will always answer politely.”

Using AWS, Jacaranda has the reliability, affordability, and scalability it needs to expand its solution. Already, PROMPTS receives 10,000–12,000 questions each day—and Jacaranda is continually improving its services to better help additional expectant parents. For example, Jacaranda uses AI to send tailored messages to its users according to their medical histories, locations, and socioeconomic status. The organization is also expanding from SMS to messaging apps and voice channels. “AWS has resources and expertise that I couldn’t even begin to dream of,” says Patel. “And because of our relationship, we can tap into that expertise. This informs our programs and how we think about them, which then impacts lives on the ground.”

PROMPTS has reached 3 million mothers across Kenya and has had a proven impact on care-seeking behaviors. For example, Jacaranda has noted a 20 percent boost in attendance at prenatal health visits and a 1.85 times increase in postpartum family planning. Jacaranda involves humans in all AI-driven tasks to maintain safety and accuracy. Questions that are flagged by its model as being of high priority are triaged to a clinical nurse, who calls the mother directly for follow-up and referral options. In one case, PROMPTS flagged a question about severe headaches and blurred vision from a mother with a history of stillbirth as urgent. Help desk agents reached out and encouraged her to seek immediate medical attention. She went to the hospital right away, and a few weeks later, she gave birth to a healthy baby.