Eddiegah/nexora-research

Fairness-aware machine learning research for preventive healthcare in sub-Saharan Africa --- > *"Only 8 of 47 clinical AI studies reviewed included any fairness evaluation.* > *Only 12 were conducted on African patient data.* > *This research program exists to change that."* I am **Edmund Eric Gah**, an AI researcher from Ghana building **fairness-aware machine learning systems** for early disease detection in underserved communities across sub-Saharan Africa. Clinical AI has a blind spot. Tools that perform well on Western populations routinely fail the communities that need them most — and almost no one is measuring it. My work addresses this gap head-on: auditing algorithmic bias, building interpretable models, and publishing evidence that makes equitable deployment impossible to ignore. **This is the Nexora AI for African Health Research Program.** Three papers in. Many more to come. --- ## 🔬 Research Pillars **⚖️ Algorithmic Fairness** Auditing ML performance disparities across sex, age, BMI, and cholesterol subgroups — before deployment, not after harm. **🔍 Explainable AI** SHAP-based feature attribution that makes model predictions clinically interpretable — not black boxes that clinicians cannot trust. **🫀 Preventive Detection** Shifting the intervention point from crisis care to early warning for diabetes, CVD, and beyond. **🌍 Low-Resource Deployment** Designing for the real constraints of clinical settings in emerging economies — not idealized Western labs. --- ## 📑 Published Papers ### `03`   Systematic Review · Algorithmic Fairness in Clinical ML **Algorithmic Fairness in Clinical Machine Learning for Non-Communicable Disease Prediction in Sub-Saharan Africa:** *A Systematic Review, Empirical Analysis, and Research Agenda*  `May 2026`  `PRISMA Protocol`  `47 Studies Reviewed (2015–2026)` | Finding | Statistic | |:---|:---| | Studies with any fairness evaluation | **8 / 47 — only 17.0%** | | Studies using formal fairness metrics | **3 / 47 — only 6.4%** | | Studies conducted on …