consulting

AI and data science consulting services — strategy, workshops, and applied research.

Senior consultant in artificial intelligence, data science & digital transformation

I take consulting engagements at the intersection of topology, data science, and artificial intelligence, with most of my work anchored in African academic and industry contexts.

My background is pure mathematics (PhD in topology) plus applied machine learning, which means I can sit on either side of a project: writing the theory and writing the production code. The engagements I take usually involve a team that has data and a real question, and either no clear method, the wrong one, or a method nobody has audited.


Services

AI strategy & architecture

Help organizations define their AI roadmap, evaluate tools and platforms, and design scalable ML pipelines tailored to their data and domain.

TDA-powered analytics

Apply Topological Data Analysis to extract insights from complex, high-dimensional datasets — shape analysis, anomaly detection, and feature engineering using persistent homology.

Deep RL model development

Design and implement Deep Reinforcement Learning solutions for resource allocation, optimization, and decision-making systems.

ML workshops & training

Hands-on machine learning workshops for universities, research institutions, and companies. Topics: TDA, geometric deep learning, Python/R for data science.


Capabilities

Four pillars, in the vocabulary advanced-analytics consultant briefs tend to use. Each card points to evidence already on this site or in the public portfolio.

Advanced analytics & modeling

Predictive and prescriptive modeling on real public data. Time-series (SARIMA, state-space models, GBM), GLMs (Poisson, Gamma, Tweedie), survival (Cox PH, Weibull AFT), classification (XGBoost), reinforcement learning (Q-learning, LP-based policies). Validation by rolling-origin backtest, calibration plots, Gini / top-decile lift.

Evidence: 13-project portfolio spanning energy (PJM, NASA POWER), insurance (freMTPL2), telecom (MTN), health logistics (Kenya KMPDC), and retail (M5).

Visualization & reporting

Bilingual (EN/FR) reporting in two formats: executable Jupyter notebooks for technical readers, long-form case-study pages for everyone else. A Streamlit app doubles the static portfolio. Python visualization (matplotlib, seaborn) plus editorial static-site design with OG cards, hreflang alternates, and structured data.

Evidence: 5 long-form case studies + Streamlit twin.

AI development & automation

Applied AI systems built end to end. Research side: topological pipelines for early-warning detection (epidemic surveillance via persistent homology), RL policies for resource allocation, geometric and equivariant deep learning. Engineering side: typed Python, FastAPI services, reproducible Docker builds.

Evidence: The Shape of Data (No Starch, 2024), TDA + RL research on projects and publications.

Collaboration & governance

Research across mathematics, engineering, and business teams. MSc and PhD supervision at AIMS, the AI.Technipreneurs workshop series for African institutions, ACAS mentoring, 20+ co-authored peer-reviewed papers. Used to translating between data engineers, AI scientists, and the people who have to act on what the model says.

Evidence: teaching & workshops, 150+ publications, h-index 12.


Industry focus

I work across sectors. The constant: I care more about whether the model is right than whether it’s impressive. Sectors I take projects in:

Health

Disease modelling, clinical trial analytics, epidemiological forecasting, health data pipelines

Finance & Banking

Credit scoring, risk modelling, fraud detection, portfolio optimization, synthetic financial data

Insurance

Actuarial modelling, claims prediction, pricing optimization, customer segmentation

Energy

Demand forecasting, grid optimization, renewable energy planning, anomaly detection in smart meters

Agriculture

Crop yield prediction, satellite image analysis, supply chain optimization, climate risk modelling

Retail

Demand forecasting, recommendation systems, customer analytics, inventory optimization

Telecom & IT

Network optimization, churn prediction, intelligent resource allocation, anomaly detection

Education & research

Capacity building, ML curriculum design, research partnerships, data literacy programmes


How I work

1
Discovery

Understand your data, domain, and objectives through an initial consultation.

2
Proposal

Design a tailored approach with clear milestones, deliverables, and timeline.

3
Execution

Build, iterate, validate. Weekly check-ins; code and notebooks shared as they're written.

4
Knowledge transfer

Deliver documentation, training, and handoff to ensure your team can maintain and extend the work.


Partners & affiliations

I’ve worked with and through the following organizations:


Projects in action

A selection of end-to-end data science projects across African open data, each with full Python code, Jupyter notebooks, and reproducible results.

Sales forecasting dashboard

Rwanda Health Supply Chain — Designed a time series forecasting model (ARIMA + Prophet, MAPE 2.8%) and built an interactive Streamlit dashboard for public health supply chain teams. Enables proactive inventory decisions and demand forecasting across departments.

Impact: Reduced stockout incidents through 2.8% MAPE demand forecasting

Churn prediction & retention

Nigeria Telecom — Built a logistic regression + XGBoost pipeline (AUC 0.91) to identify at-risk telecom customers — analyzing usage patterns, tenure, and billing signals. Contributed to a 15% improvement in customer retention.

Impact: AUC 0.91 — 15% improvement in customer retention

Resource allocation via RL

Kenya Health Facilities — Q-learning system for mobile clinic scheduling in underserved regions. Extended healthcare coverage by 20 percentage points over random allocation.

Impact: +20pp healthcare coverage over baseline allocation

Automated billing from images

South Africa Logistics — Built a computer vision pipeline (OpenCV + Tesseract OCR) to extract parcel dimensions and text, generating itemized shipping bills automatically. Reduced manual entry time by 70%.

Impact: 70% reduction in manual data entry time

Epidemiological surveillance

Benin (INSAE / WHO-AFRO) — Built a topological anomaly detection pipeline for early epidemic outbreak detection using persistent homology on spatiotemporal case distributions. Detects cluster formation 2–3 weeks before classical WHO alert thresholds.

Impact: Epidemic detection 2-3 weeks ahead of WHO classical thresholds

Credit risk scoring

South Africa / Kaggle — Designed an XGBoost credit scoring model (AUC 0.91) using mobile money and geographic features for unbanked populations. Includes fairness analysis across demographics.

Impact: AUC 0.91 — fairness-audited credit scoring for unbanked populations

Technologies: Python (Pandas, NumPy, SciPy, scikit-learn, XGBoost, Statsmodels), SQL, Streamlit, OpenCV, PyTorch, Prophet, Reinforcement Learning


Get in touch

If you have a project that needs the kind of work described above, send me an email. I’m happy to have a short call to see whether it’s a fit before either of us commits anything.

Or connect via GitHub · Google Scholar · ORCID