Our Research
Advancing the frontiers of AI in healthcare through rigorous research focused on Africa's unique challenges and opportunities, with deep roots in Kenya.
Research Focus Areas
Language Adaptation
Developing techniques to adapt AI models to understand and process diverse African languages, enabling more natural and effective healthcare communication across the continent.
Privacy-Preserving AI
Researching federated learning and other techniques that enable AI model training without centralizing sensitive patient data.
Cultural Context in AI
Exploring how diverse cultural contexts across Africa affect AI performance in healthcare settings and developing methods to create culturally-adaptive AI systems that respect local traditions.
Low-Resource Computing
Developing techniques to optimize AI models for low-bandwidth environments and resource-constrained devices common in rural healthcare settings.
Medical Imaging AI
Researching computer vision techniques for medical imaging analysis, with a focus on conditions prevalent across Africa and images from diverse skin tones representing the continent's rich diversity.
AI Ethics & Governance
Developing frameworks for ethical AI deployment in healthcare settings, with a focus on fairness, transparency, and accountability across diverse African contexts while honoring our Kenyan heritage.
Recent Publications
Adapting Medical AI Models to African Languages: Challenges and Solutions
Dr. Sarah Mwangi, James Omondi, Dr. Thomas Okello
Journal of Healthcare Informatics • October 2023
This paper explores the challenges of adapting medical AI models to understand and process diverse African languages, with a focus on Swahili, Amharic, Yoruba, Zulu, and Hausa. We present novel techniques for transfer learning and data augmentation that significantly improve model performance in low-resource language settings across the continent.
Federated Learning for Privacy-Preserving Healthcare AI in Rural Settings
James Omondi, Dr. Amina Kimani, Michael Odhiambo
AI for Global Health Conference Proceedings • August 2023
We present a novel federated learning approach that enables AI model training across distributed healthcare facilities in rural Africa without centralizing sensitive patient data. Our approach addresses challenges of intermittent connectivity and device heterogeneity while maintaining model accuracy across diverse regional settings.
Cultural Context in Medical AI: Improving Diagnostic Accuracy Through Cultural Adaptation
Dr. Amina Kimani, Grace Achieng, Prof. David Otieno
Ethics in Artificial Intelligence • June 2023
This study examines how incorporating cultural context into medical AI systems affects diagnostic accuracy and patient trust. Through a series of controlled experiments in diverse African communities from Kenya to Nigeria to South Africa, we demonstrate that culturally-adapted AI systems achieve significantly higher accuracy and user acceptance across the continent's varied cultural landscapes.
Low-Bandwidth Optimization Techniques for Healthcare AI in Remote Areas
Daniel Muthoni, Thomas Okello, Dr. Sarah Mwangi
IEEE Transactions on Medical Imaging • April 2023
We present a suite of optimization techniques that enable sophisticated medical imaging AI to function effectively in low-bandwidth environments. Our approach combines model compression, progressive loading, and adaptive resolution to deliver diagnostic assistance even in areas with limited connectivity.
Our Research Partners
University of Nairobi
Kenya
Makerere University
Uganda
University of Dar es Salaam
Tanzania
MIT Media Lab
USA
Stanford HAI
USA
DeepMind Health
UK
Join Our Research Efforts
We're always looking for research partners, collaborators, and talented researchers to join our team.