Research Papers
My published research, conference papers, and academic work
LexAfriq: A Multimodal AI System for Early Dyslexia Screening in African Classrooms
We present LexAfriq, a mobile-first dyslexia screening system designed for low-resource African contexts. The tool leverages multimodal machine learning — combining handwriting analysis, eye tracking, and gamified cognitive tasks — to identify early signs of dyslexia in children without the need for expensive equipment or clinical supervision. Evaluated on simulated classroom data, LexAfriq achieves high accuracy while maintaining accessibility and child-friendly design. This work contributes a context-aware framework for equitable AI deployment in education.
Retrieval-Augmented Question Answering over Ghanaian Public Health Policies
This report presents a Retrieval-Augmented Generation (RAG) pipeline for answering questions grounded in Ghana’s Ministry of Health policy documents. Using 31 real-world PDFs, we demonstrate that combining dense retrieval with a flan-T5 generator improves factual accuracy by 33% and eliminates hallucinations compared to baseline language models.