Background
This project presents Saytů Hemophilie, a novel AI-powered educational chatbot developed for People with Haemophilia (PWH) and their families in Senegal. Available on Android and iOS in both French and Wolof, the chatbot addresses critical health literacy gaps by providing accessible, reliable information on symptom management, treatment, and risk prevention. It serves as a supplementary resource to clinical care, supporting users in managing their condition with greater confidence and autonomy.
In Senegal, limited awareness of haemophilia contributes to poor outcomes, stigmatization, and barriers to care. Saytů Hemophilie empowers patients and caregivers by bridging knowledge gaps through culturally and linguistically adapted content. A central innovation is its use of Wolof, a major national language, which enhances inclusivity especially for women carriers and family members with limited formal education.
Objectives
This project seeks to address this gap by exploring how LLMs can be adapted and deployed inlow-resource settings to enhance education and awareness about haemophilia in local languages.
Additionally, it will investigate the integration of machine translation systems into LLMs to make healthcare information accessible to linguistically diverse African populations, ultimately improving diagnosis, access to treatment, and overall healthcare capacity. The low-resource African languages of interest in this project are Wolof and Swahili. Since this project would be piloted in countries across Sub-Saharan Africa and Eastern Africa, it makes sense to select languages that are widely spoken in these 2 regions. For this pilot, we are planning on deploying our LLM in Senegal, where approximately 80% of the population speak Wolof.
Methods
This project will evaluate the effectiveness of large language models (LLMs) as educational tools for haemophilia through a multi-phase approach which involves first fine-tuning a medically trained LLM with haemophilia-specific information sourced from peer-reviewed research, open-source medical books, and validated clinical guidelines. The resulting haemophilia-focused model will then undergo rigorous evaluation and refinement through testing with medical professionals and patients. Finally, this optimized LLM will be integrated into a mobile application, enabling users to access accurate and accessible educational support.
Keywords: Education; haemophilia; LLM; QoL; Africa





