RESEARCH PROJECTS
Exploring the frontiers of computational intelligence through real-world applications and groundbreaking research implementations.
A real-time epidemic tracking platform using swarm-optimised agent-based models, piloted across 5 regions of Ghana with the Ghana Health Service. Generates 14-day disease spread forecasts 40% more accurate than traditional models.
Connects 12 district hospitals in a federated learning network, enabling joint training of disease prediction models on EHR data without raw patient data leaving each hospital.
Reinforcement learning agents that dynamically re-route vaccine distribution during supply disruptions, reducing cold-chain wastage by an estimated 23% in Ghana's Northern Region.
A proof-of-concept authentication system using zero-knowledge succinct proofs to verify fingerprint and iris matches without ever exposing raw biometric templates to a server.
End-to-end platform ingesting multi-spectral drone imagery and soil sensor data to map crop stress, nutrient deficiency, and yield potential at 10 cm resolution across smallholder plots.
Ensemble learning framework combining satellite phenology, weather station data, and farmer survey inputs to produce district-level crop yield forecasts for staple crops across 8 African countries.
A family of quantised YOLO-variant models and inference runtimes enabling real-time object detection on Raspberry Pi 4 and similar edge hardware for wildlife monitoring and agricultural scouting.
NeRF and Gaussian-splatting pipeline for rapid, photorealistic 3D scene reconstruction from smartphone video, with applications in cultural heritage digitisation and remote infrastructure inspection.
Meta-learning framework enabling rapid fine-tuning of large foundation models to new health and agricultural tasks with as few as 5–10 labelled examples per class.
Toolchain for automated bias detection, causal attribution, and counterfactual explanation generation for ML models deployed in high-stakes African institutional contexts.
Decentralised swarm of 20+ UAVs that coordinate search-and-rescue operations in GPS-denied post-disaster environments using only local communication and stigmergic shared maps.
Swarm-optimised demand-response and fault-isolation agents deployed on a microgrid testbed at the University of Ghana, reducing peak load variance by 31%.