Research & Publications
Currently pursuing my Master in Engineering Artificial Intelligence at Carnegie Mellon University Africa. My research is centered on building intelligent, efficient, and safe systems inspired by human learning, low-resource NLP, and trustworthy AI.
Building Intelligent & Safe Machines
My research interest is building intelligent and safe machines by borrowing the most fascinating ideas from the brain and the world around us. I have always been passionate about learning, and I have tried many techniques to learn more efficiently.
My core interests come from two experiences. The first is spaced repetition, which I use myself. It hands the question of when to revise over to a learning algorithm that predicts when I am about to forget a concept and brings it back at that moment. I find it fascinating that a simple model of human memory can make learning this much more efficient. The second is my earlier work in NLP, particularly Named Entity Recognition for low-resource languages, which led to my co-authored paper on LinguoNER. Together, these shape my interest in systems that learn the way humans do: efficiently, with structure, and with limited data.
More recently, I have started reading about AI safety, especially the impact of uncontrolled, unexplainable black-box models across different areas of life. This interest is still in its exploration phase and I have no research results in it yet, but it builds on the others. Systems that learn efficiently must also be understandable and trustworthy, otherwise their benefits come with risks we cannot see. Just as a model’s weights are updated during backpropagation, I keep updating myself, and this is the direction my learning is currently taking.
Research Domains
Core focus areas guiding current academic investigations and thesis work.
Cognitive & Memory-Inspired Learning
Investigating how models of human memory, spaced repetition, and structured knowledge retention can make machine learning more sample-efficient.
Low-Resource African NLP
Language-agnostic Named Entity Recognition, syntactic bootstrapping, and corpus generation for under-represented Bantu and Cameroonian languages.
AI Safety & Trustworthy Systems
Exploring explainability, governance, and interpretability to understand the risks of black-box models and ensure efficient systems remain trustworthy.
Multimodal Speech & Language Preservation
Cross-lingual audio-visual alignment, isometric translation architectures, and neural voice synthesis to bridge digital access barriers in indigenous African languages.
Peer-Reviewed Publications
Archival journal articles and conference proceedings.
Informatics 2026, 13(2), 31
Addresses severe data scarcity in Named Entity Recognition (NER) for extremely low-resource languages by proposing LinguoNER, an extensible, language-agnostic bootstrapping framework. Evaluated on Yambeta (a Bantu language spoken in Cameroon), the framework demonstrates effective transfer learning and annotation strategies to preserve cultural and linguistic heritage while empowering downstream African language technologies.
Interested in Collaborating on AI Research?
I am always eager to collaborate with researchers, labs, and practitioners interested in African NLP, multimodal dubbing, or speech technologies.