When the Stars of Innovation US-Kenya AI Challenge announced its finalists, for a breathless second, the room stopped being a conference hall and became a shared heartbeat.

Lights narrowed to the stage, a slide flashed the MediBora logo, and five students — equal parts exhaustion and elation — rose as if the prize had already begun to rewrite the rest of their year.

The Kenyan students met at the hackathon. What began as a meeting of strangers became a team because each member carried a distinct way of seeing the same problem: Pregnant women were slipping through the last mile.

“As we started talking, we realised we all came from different backgrounds but shared an interest in healthcare, and that maternal health was a problem that needed to be handled with seriousness,” Jael Wainaina, the team’s data scientist, told bird.

“Maternal health is a real and growing problem in Kenya and even globally, and we felt it needs to be handled with the seriousness and care it deserves.”

According to the World Health Organisation, Kenya records an estimated 355 maternal deaths per 100,000 live births, a rate that remains far above the global target of fewer than 70 deaths by 2030. While policy changes have expanded access to maternal care, pregnancy remains one of the country’s most persistent public health risks.

When Wainaina’s colleague, Fardosa Mohamed, speaks during the call, she does so quickly, like someone who has been rehearsing the argument long before anyone asked.

“We all came in as strangers through the competition, but once we started sharing stories about maternal health, it became clear we were looking at the same problem from different angles,” she said. “That is what brought us together.”

Across the continent, the pattern is starker. According to Unicef, Africa accounts for about 70 per cent of global maternal deaths, with a regional maternal mortality ratio of roughly 454 deaths per 100,000 live births. Many of those deaths occur outside hospitals, after symptoms emerge but before help arrives.

They pooled their skills — developers, a biomedical engineer, business minds and data specialists — and named what they would build: MediBora, a maternal health platform designed to flag early pregnancy risks and shorten the time between symptoms and care. 

HOW IT WORKS

MediBora combines AI-assisted risk logic with communication channels already common in Kenyan households, using mobile apps where possible, alongside SMS, USSD and voice to link mothers, clinicians and community health workers into a single digital loop.

That design reflects where breakdowns often occur. According to the World Bank, delays in deciding to seek care contribute to roughly one-third of maternal deaths in Kenya, while about 70 per cent of pregnant women live in rural areas, where distance, cost and transport gaps slow emergency referrals.

“We realised there is a big gap because many mothers skip clinic visits, so hospitals are not able to track their progress or detect early warning signs while they are at home,” Wainaina said.

That gap shaped every technical decision the team made.

MediBora relies on routine check-ins delivered through text messages or voice prompts, asking mothers to report symptoms such as headaches, swelling, dizziness or reduced foetal movement — signals clinicians already recognise as early warning signs. Those responses feed into a rules-based risk scoring system built from established clinical guidance rather than live patient experimentation.

“We are very careful about this,” Wainaina said. “At this stage, we have not tested on patients. It is still a prototype, and we are working on approvals so that validation can happen ethically and safely.”

The system aggregates reported symptoms and available indicators, flags elevated risk and escalates that signal to clinicians. From there, human-led systems take over.

“There is a patient dashboard and a doctor’s dashboard,” Mohamed said. “If a doctor sees abnormal vitals or warning signs, they can call the mother in early or initiate referral before it becomes an emergency.”

In higher-risk scenarios, the platform is designed to trigger GPS-enabled emergency alerts, allowing transport to be arranged sooner and referral facilities to receive advance notice. That design reflects existing communication habits.

According to the Communications Authority of Kenya, the country had about 75 million mobile connections by late 2025, representing more than 140 per cent penetration, with SMS access extending even to households without smartphones.

“That digital file matters,” Mohamed said. “In referral hospitals, nurses often get no advance notice, and minutes or hours of delay can decide survival when haemorrhage or hypertensive emergencies strike.”

REASSURING MUMS

On the project’s website, the ambition is listed under an “Our Impact” banner: SMS health alerts, personalised care, emergency support, early risk detection, comprehensive monitoring and enhanced provider support.

Those features mirror known system gaps. According to Jacaranda Health, its AI-enabled Prompts platform has reached more than three million Kenyan women, with 85 per cent of high-risk users seeking facility care after receiving alerts, suggesting that timely digital signals can influence care-seeking behaviour.

The leap from podium to ward, however, is less technical than institutional.

Mobile health tools often sit in a procurement grey zone. Until ministries and hospitals define clear evaluation and purchasing pathways, many pilots stall after proof of concept.

“You can build something brilliant,” Mohamed said, “but scaling it requires partnerships, funding and people willing to invest and believe in it, especially when you are students.

“We will measure how many risk alerts we detect, how fast facilities respond and how many successful referrals we enable.”

For the group, trust matters, too.

“If mothers use the system and recommend it, that is real impact. We want mothers to feel reassured every day, without unnecessary hospital visits, stress or cost.”

The team is candid about what it had to unlearn.

“One is assuming that what works for one mother will work for another,” Wainaina said. “Symptoms that are normal for one person can be dangerous for someone else.”

That insight shaped MediBora’s architecture. Symptom thresholds vary. Nutrition advice varies. Communication preferences vary. Phone access varies. The platform’s multi-channel design reflects lived conditions, not branding logic.

“This project pushed me to think like a health professional and a problem solver, not just a data scientist,” Wainaina said.

They are careful about how they describe what they are building.

“We are not trying to replace clinics,” Mohamed said. “We want to work with them and support healthcare workers to save lives.”

What comes next is procedural but decisive: pilot permissions, legal guidance, data protection frameworks, interoperability standards and clinics willing to document outcomes.