It is unacceptable that in 2019, we can do our communications, our banking, our navigation, our information searches through a single device…and still nearly 1 million children die of a perfectly treatable disease, pneumonia, that gets diagnosed too late. This talk dives into the experience Dr Elina Naydenova, a biomedical & ML engineer, has had in developing and delivering AI-powered diagnostic solutions for low-resource settings. Elina is the co-founder of Feebris, a health tech start-up whose ML-powered health platform enables minimally trained health workers and carers to detect complex conditions in the field. The platform uses advanced ML to process noisy signals from point-of-care devices and detect early markers of disease. Trained against a robust clinical evidence base, the algorithms have been shown to predict the presence of conditions like pneumonia with high accuracy. To deliver this innovation in low-resource settings, the Feebris team is tackling the challenges of unreliable connectivity and generalisability cross-populations. This talk will discuss both the opportunities and challenges in developing and delivering such solutions for the most vulnerable populations.