Esther Mulu, 69, supports her seven children and 25 grandchildren on a small farm in Kitui County, Kenya – one of six counties facing recurring droughts and shortened rain seasons. With no other income and dwindling food aid, her family now survives on a single daily meal, often just porridge.
“If you can invent a technology to end hunger, I’ll be forever grateful,” Mulu says.
That’s exactly what researchers at the University of California, with support from Microsoft, are working on. They’re developing an AI-driven tool that can predict food insecurity months in advance using a decade’s worth of data, including satellite imagery, weather trends, and clinical health records.
The tool, tested since early 2023 in drought-prone regions, helps forecast malnutrition cases 3 to 6 months ahead. AMREF’s Samuel Mburu says this early warning system allows counties to better allocate resources, especially for children under five.
“We now have the ability to use data not just for reporting, but for prediction—and that’s critical for saving lives,” Mburu explains. He adds that the model, once focused on six counties, now covers all of Kenya. Future updates aim to include community-level data for even better accuracy.
The main challenge? Cost. While Microsoft’s sponsorship helped launch the pilot, scaling cloud-based AI solutions remains expensive.
Still, AMREF is optimistic. With predictive power, local health officials can act earlier—mobilizing food packs and medical aid before hunger reaches crisis levels.
