A Kenya–Michigan Partnership Is Redefining Global Health Security

A health worker in a Kenyan border county notices something unusual: several patients, all bleeding in ways that don't fit anything routine. Word travels the way it usually does at first — a phone call, a WhatsApp message, a conversation between neighbors before it becomes a report. Two days pass before anything is written down. A third day passes before anyone decides whether the national government needs to know. By the time a response team arrives, the same symptoms have turned up in the next county over.

Nobody missed the warning signs. The problem was speed: the space between noticing and knowing what to do about it. In an outbreak, that space is where things go wrong.

Kenya's national public health institute has spent the past several years trying to close it, with an unusual set of partners: a UK-funded programme aimed at strengthening disease response across Africa, a center focused on pandemic intelligence, and artificial intelligence researchers at the University of Michigan. What they built together is not, on its face, a piece of software. It's a decision-making framework: a structured way of asking what an event is, how serious it is, and who needs to know. Only once that framework had been tested, reviewed, and validated by Kenyan epidemiologists and emergency responders did anyone consider handing part of it to a machine.

"Public health guidance existed, but decision-makers lacked a practical tool to consistently translate surveillance information into timely action," says Mark Nanyingi, PhD, the epidemiologist who led its technical development. The framework took shape through national workshops, simulations, and stress-tests against Kenya's hardest scenarios: Ebola-like hemorrhagic fevers, cholera, chemical spills, outbreaks with no clear cause at all.

The team didn't bring in AI until that groundwork was solid. Geoffrey Siwo, PhD, who leads AI work at Michigan's Center for Global Health Equity, and his data science team built an AI Agent trained not to make judgment calls, but to apply the same logic a trained epidemiologist would, instantly, every time.

Before anyone trusted it with anything real, they tested it: side-by-side comparisons between the AI's recommendations and those of human experts, across scenario after scenario. The test results were close enough that the team is now confident the system can hold up under pressure, a conclusion detailed in a preprint the group published on the work.

The recent Ebola outbreak in East Africa was a reminder of the stakes. Early signals never arrive whole. They show up as fragments — a rumor, a strange symptom, a handful of cases that might be nothing — and someone has to decide, fast, what they add up to. That is the exact moment the framework and its AI Agent are built for: not replacing the epidemiologist in the room, but giving them a faster, steadier way to think through what they're seeing.

"The real innovation is not artificial intelligence itself," Nanyingi reflects. "It is the combination of public health expertise, operational experience, and technology working together to help decision-makers act faster and more confidently when lives are at stake."

That, more than the technology, is the part worth paying attention to. AI is arriving in global health faster than most institutions can evaluate it, often built far from the places it's meant to serve. This project took the slower route: years of Kenyan-led framework-building before a line of AI code was written, and a rigorous evaluation before anyone called it ready. It's a quieter story than "AI predicts the next pandemic", but it may be the more honest version of what responsible AI in public health actually looks like: not a shortcut, but a tool built patiently, by the people who'll be the ones relying on it when the next warning sign appears.

Project Team

Mark Nanyingi
Center for Global Health and Pandemic Intelligence (CGP), Kenya

Eric Osoro
Washington State University Global Health Kenya (WSU GHK), Kenya
Paul G Allen School for Global Health, Washington State University, USA
Center for Global Health Research, Kenya Medical Research Institute (CGHR-KEMRI), Kenya

Geoffrey H. Siwo
Department of Learning Health Sciences, University of Michigan Medical School, USA
Department of Pharmacology, University of Michigan Medical School, USA

Isaac Ngere
Washington State University Global Health Kenya (WSU GHK), Kenya
Paul G Allen School for Global Health, Washington State University, USA
Center for Global Health Research, Kenya Medical Research Institute (CGHR-KEMRI), Kenya

Samuel Kadivane
Kenya National Public Health Institute (KNPHI), Ministry of Health, Kenya

James Magige
Center for Global Health and Pandemic Intelligence (CGP), Kenya

Joseph Kamau
One Health Centre, Kenya Institute of Primate Research (KIPRE), Kenya

Shreya Jain
Department of Learning Health Sciences, University of Michigan Medical School, USA

Bryan O. Nyawanda
Center for Global Health Research, Kenya Medical Research Institute (CGHR-KEMRI), Kenya

Joseph Wanyoike
Center for Global Health Equity, University of Michigan, USA

Ian Njeru
Center for Global Health and Pandemic Intelligence (CGP), Kenya

Kadondi Kasera
Tackling Deadly Diseases in Africa Programme II, Palladium, Kenya

Victoria Kanana
Kenya National Public Health Institute (KNPHI), Ministry of Health, Kenya

Kamene Kimenye
Kenya National Public Health Institute (KNPHI), Ministry of Health, Kenya