FarajaMH: Culturally Adapted Generative AI Model for Mental Health Screening in Kenya and Tanzania; Leveraging Longitudinal Data, Clinical Notes and Conversational Chats (360G-Wellcome-333073_Z_25_Z)

This project aims to build, validate, and ethically evaluate a bilingual (English and Swahili) generative AI (FarajaMH) model for screening of depression, anxiety, and psychosis (DAP) through natural language interactions. The project combines research and lived experiences to address the lack of culturally attuned tools for early screening and referral in Kenya and Tanzania. The word Faraja means “comfort” in Swahili, symbolising the project’s goal of creating a compassionate GenAI model that will recognize DAP, expressed in everyday metaphors. FarajaMH is a model trained with mental health conversational chats, clinical notes, audio and textual screening data from Health Demographic and Surveillance Systems (HDSS), for screening DAP from sampled populations in Kenya and Tanzania. It will be evaluated and validated entirely within a controlled research environment, testing its screening accuracy, usability, and safety. The expected outcomes include (1) a harmonised multimodal data repository of mental health information mapped to the Observational Medical Outcomes Partnership Common Data Model (OMOP CDM); (2) a bilingual generative AI model for screening for DAP; (3) a mental-health data catalogue and visual analytics dashboard; and (4) publications and conference presentations on GenAI in mental health in East-African context.