Research Products
AI Tools for Health Messaging, Inclusion and Evaluation
We turn our research questions into real-world tools – AI models, data pipelines, infrastructure, and evaluation systems – designed to serve public health systems, innovators, and researchers across Africa.
InfoAFYA™ LLM
A behavior-first language model for real-time health messaging in low-resource settings.
The InfoAFYA COM-B LLM is a fine-tuned language model designed to generate empathetic, actionable health messages using the COM-B behavior change framework. It powers the InfoAFYA chatbot and supports health promotion campaigns that address real-life barriers caregivers and youth face and currently supports Sickle Cell Disease communities.
The InfoAFYA COM-B LLM is a fine-tuned language model designed to generate empathetic, actionable health messages using the COM-B behavior change framework. It powers the InfoAFYA chatbot and supports health promotion campaigns that address real-life barriers caregivers and youth face and currently supports Sickle Cell Disease communities.
What makes it unique?
-
Built on a fine-tuned LLaMA 3.1 8B model with over 1,000 examples grounded in COM-B and local realities
-
Trained for behavioral tone, coaching flow, and empathy, not just information delivery.
-
Integrated to RAG pipelines and guardrails for safe, grounded answers.
-
Outputs include both a health promotion message and a behavioral challenge statement, among other contents.
-
First COM-B-aligned LLM tested in a live WhatsApp/SMS health assistant.
-
Supports multi-turn health education, contextual nudging, and open-ended Q&A.
-
Acts as the behavioral reasoning engine within InfoAFYA™ Plus solution.
Data Services & Collaborations
We offer end-to-end support for ethical data collection, annotation, validation, and model tuning. We champion equity by designing with marginalized voices, prioritizing linguistic inclusion, open evaluation, and community-led innovation. All our outputs, from tools to data, are shared as digital public goods, grounded in transparency, and accessibility for all.