FarajaMH
Culturally Adapted Generative AI Model for Mental Health Screening in Kenya and Tanzania
Tatenda Kavu · African Population & Health Research Centre · Kenya
What they're working onFarajaMH is a project developing a culturally adapted generative AI model to improve the early screening of depression, anxiety, and psychosis in Kenya and Tanzania. The project addresses the limitations of existing mental health screening tools, which often fail to capture culturally specific expressions of psychological distress in English and Swahili. By leveraging longitudinal mental health data, clinical notes, and conversational data from health facilities and communities, FarajaMH will build and evaluate a bilingual large language model designed for primary healthcare settings. The project integrates data harmonization using international standards, culturally grounded symptom detection, large language model fine-tuning, and rigorous evaluation for clinical accuracy, safety, usability, and fairness. Outputs will include a FAIR-compliant multimodal mental health data repository, a validated bilingual large language model, a scalable AI-assisted screening application, a mental health data catalogue, and analytics dashboard, creating opportunities for equitable, AI-enabled mental healthcare across East Africa.
AI approach
Generative AI · bilingual LLM · multimodal/longitudinal data
Mental health focus
Depression · anxiety · psychosis · screening
Social media