Building intelligent and explainable AI systems for healthcare, computer vision, and predictive analytics.
Data Scientist — Deep Learning & Explainable AI
I build interpretable computer vision and predictive models across healthcare, finance, and industrial applications — from tuberculosis screening to automated damage assessment. My latest work compares three CNN architectures for TB classification from chest X-rays, reaching 99.04% accuracy with Grad-CAM++ explainability — completed under the DATICAN / NIH DS-I Africa scholarship.
I'm a Data Scientist and AI Engineer with a strong foundation in Machine Learning, Deep Learning, Computer Vision, and Explainable AI. I recently completed an MSc in Computer Science under the DATICAN Data Science Scholarship, a competitive NIH-funded programme, where my research compared deep learning models for tuberculosis classification from chest X-ray images using Grad-CAM++ explainability.
I've applied this foundation to projects across healthcare, finance, and predictive analytics using Python, PyTorch, and Scikit-learn, and I bring over three years of enterprise IT and cloud support experience — a background that shapes how I think about deploying and maintaining models reliably in production.
I'm currently looking for a data scientist role where explainability isn't an afterthought — in health-tech, fintech, or with organisations that need models people can actually trust and act on.
What I work with
A toolkit built for taking a model from raw data to a deployed, explainable result.
Programming & Libraries
Machine Learning & AI
Tools & Platforms
Selected work
Six projects spanning medical imaging, finance, and applied computer vision.
Explainable Tuberculosis Classification
PyTorch · DenseNet121 · ResNet50 · EfficientNet-B0 · Grad-CAM++
- Developed and compared three CNN architectures for binary TB classification across five combined chest X-ray datasets (~10,400 images)
- Applied Grad-CAM++ to generate clinically interpretable heatmaps, confirming upper-lobe activation in TB-positive cases
- Reached 99.04% accuracy and an AUC-ROC of 0.9990 with DenseNet121 under a class-weighted loss strategy
Credit Risk Prediction Model
A classification model predicting loan default risk from customer financial and demographic features, evaluated on accuracy, precision, recall, F1, and AUC-ROC.
Insurance Premium Prediction
A regression model estimating insurance premiums from health and lifestyle features, with eight algorithms benchmarked and the best performer deployed.
Car Damage Prediction
An image classification model that detects and categorises car damage directly from photographs.
Expense Tracking System
A personal finance tracker with data visualisation and reporting features for day-to-day spend analysis.
CodePotato Disease Classification
A computer vision model classifying potato leaf diseases from images, deployed as an interactive Streamlit application.
CodeBackground
From enterprise IT support into applied deep learning research.
Experience
- Provided enterprise-level Microsoft 365 support across EMEA, covering Exchange, Teams, SharePoint, and Azure AD
- Implemented security controls and compliance configurations for enterprise clients
- Mentored junior support engineers, two of whom reached tenure status
- Consistently resolved complex escalation cases within SLA targets
- Provided day-to-day IT support across hardware, software, and network infrastructure
- Managed user accounts, system configurations, and internal IT documentation
- Supported business continuity by minimising system downtime across departments
Education
- Research: comparative analysis of deep learning models for TB classification from chest X-rays, with explainability
- Combined five public datasets (10,398 images) into a single training corpus
- 99.04% accuracy, 98.62% sensitivity, 99.34% specificity, AUC-ROC 0.9990 with DenseNet121
- Grad-CAM++ applied to visualise clinically plausible activation across all three CNN architectures
Ongoing development
Continuing to build depth alongside the thesis work.
DATICAN Data Science Scholarship
NIH-funded DS-I Africa initiative, 2024–2026
CodeBasics Data Science & Generative AI
Machine learning, deep learning, SQL, Python, statistics
ALX Virtual Assistant
Professional development certification
Let's talk about where explainable AI can make a difference.
Open to data scientist roles in health-tech, fintech, NGOs, and remote-first teams. Feel free to reach out directly.