Statistician and data scientist in Bauchi, Nigeria. I build uncertainty-aware models and open tools for health, poverty and inclusion data.
I have over ten years of experience in institutional statistics at Abubakar Tafawa Balewa University (ATBU), where I am now a Senior Statistician, supporting National Universities Commission (NUC) reporting, accreditation and Senate decisions.
- Bayesian and machine-learning models, always reported against an honest baseline and with their uncertainty shown.
- AI and LLM applications (RAG): retrieval measured on held-out question sets, answers that cite their sources and are checked automatically, and an AI judge used alongside human-checked ground truth. Example: DRLCA.
- End-to-end pipelines on free cloud infrastructure: scheduled ingestion, provenance manifests, tests and CI.
- Accessible apps: screen-reader-safe layouts, keyboard navigation, voice input and high contrast.
- Reproducible research: raw data to manuscript-ready tables and figures in code, with R, Python and Quarto.
- Statistical consultancy and training through JJMB Datascience Hub.
| Project | Result | Stack | Links |
|---|---|---|---|
| Price intelligence pipeline 📍 Nigeria |
XGBoost demand forecast at 10.9% MAPE vs a 13.8% naive baseline; fine-tuned FinBERT at 76.8% vs 71.0% majority class; LP optimiser for price and stock moves. Synthetic sales data, documented as such. | Python, XGBoost, FinBERT, SciPy, Supabase, GitHub Actions | repo · dashboard |
| Crude Blend Optimizer 📍 Nigeria |
ML yield surrogates (5-fold R² ≥ 0.982) drive differential evolution to within $0.005/bbl of the exact LP optimum; VaR/CVaR over 10,000 price scenarios; 140 offline tests. | Python, scikit-learn, SciPy, Docker | repo · site |
| Bayesian World Cup | Hierarchical model of 22 World Cups with partial pooling; 50,000-run Monte Carlo tournament, counterfactual do() scenarios, served from precomputed posteriors. |
Python, PyMC, Gradio | repo · app |
| Disability Rights Assistant (DRLCA) 📍 Nigeria |
Multi-turn chatbot over Nigeria's Disability Act 2018, the 1999 Constitution and a 2020 factsheet. It quotes the law word for word with a citation on every excerpt, puts the DRAC toll-free helpline first and lists 20 browser-verified disability organisations. It works offline by default; an opt-in plain-language AI answer is checked against the excerpts it cites. Measured honestly: held-out strict recall 0.63–0.65 on two blind sets, 0 false refusals, refusal calibration 0 / 212. WCAG 2.2 AA with 0 axe issues across 33 views. English, Hausa, Igbo and Yorùbá interface. It states that no lawyer has reviewed it yet. Zero cost. | Python, TF-IDF retrieval, Streamlit, Gemini (opt-in), Playwright + axe, GitHub Actions | repo · app · demo video |
| Impact Mosaic (Ayuba Gufwan disability programme) 📍 Nigeria |
Independent, source-linked showcase of the programme plus a learning sandbox: 8 courses and 48 lessons, experimental Hausa, Yoruba and Igbo journeys, administrator learner maps by state and LGA, scorecards and grant-planning reports. Zero axe accessibility violations on the audited scorecard views. | React, TypeScript, Vite, Cloudflare Pages Functions, Playwright | repo |
| Fieldmargin (crop value predictor) 📍 Nigeria |
Offline-first planner that lets Nigerian farmers compare crop margins under their own land, yield, price and cost assumptions, with custom crops and bag equivalents; installs on a phone and keeps working offline. | React, TypeScript, Cloudflare Pages | repo · app |
| Nigeria Equity Atlas 📍 Nigeria |
State-level Multidimensional Poverty Index for all 36 states and the FCT, with conflict and climate overlays and a four-round trend, from a numbered, re-runnable pipeline. | Python, Tableau | repo |
| Messi vs Ronaldo, weighted | Difficulty-weighted scoring rates (FAMD weighting) with a 10,000-resample match bootstrap; intervals and effect sizes instead of significance labels. | R, Shiny, bslib, Docker | repo · dashboard |
| Sol Searching 📍 Nigeria |
50 verified Kaduna businesses from a one-time OpenStreetMap import turned into a ranked solar sales pipeline with payback and risk–reward. | Python, Streamlit, OpenStreetMap | repo · demo |
| Urban Agriculture, Where It Concentrates 📍 Nigeria |
Spatial study of 484 OpenStreetMap cultivated-land features across 775 LGAs: Moran's I = 0.300, only 19 LGAs significant after FDR control — 13 in Kano, 6 in Lagos, none elsewhere. Density falls 66% per doubling of LGA area, and the effect strengthens (−1.001 → −1.421, ΔAIC ≈ 20) once uneven OSM mapping coverage is controlled for; a rainfall hypothesis is withdrawn rather than claimed. | R, sf, spdep, spatialreg, MASS, Quarto | repo · source · summary panel · LISA map |
- 🎓 PhD Statistics, ATBU (2026–present): in progress; hybrid Bayesian deep learning for spatio-temporal risk forecasting
- 🥇 MSc Statistics, ATBU (2024): Best in class (top 1%)
- 🥈 BSc Statistics, University of Maiduguri (2013): 2nd best in class (top 2%)
- 🏅 MTN Foundation Scholarship (2009–2012), plus a place on the Dean's List every year from 2009 to 2012
- 📜 Commendation letter from ATBU's Directorate of Academic Planning (2021)
- 🏛️ CISON / NANS — Registered Statistician, member ID 20260732, joined September 2026. All three fees paid; status and membership verified on the portal.
- Dandali: maps 51,022 health facilities live from the GRID3 ArcGIS feed and falls back per feed when one is offline. Plain JavaScript and Leaflet, no build step. Demo.
- ATBU Academic Planning Portal: a desktop app that manages 223 professor records and exports them in NUC format. Python, Flet and SQLite.
7 peer-reviewed papers in health and applied statistics, 5 citations.
Research pipelines, each reproducible from raw data to manuscript tables and figures:
- Bayesian change-points in Nigerian rainfall: a hierarchical model of 74 years of rainfall at 9 stations in 3 climate zones; LOO cross-validation rejects the no-change model in every zone. Python and PyMC; manuscript in preparation.
- Avian influenza in Nigeria, systematic review: PRISMA 2020 screening of 2,157 records from six databases down to 113 included studies.
- Anti-trypanosomal plant extracts: parses a messy Excel workbook into tidy data and runs block-level non-parametric tests with Holm adjustment through to a Word report. R and Quarto.
- Non-O157 E. coli in fermented milk: prevalence with 95% CIs, ANOVA and antibiotic-susceptibility profiles for 200 isolates.
- Urban agriculture in Nigerian cities, spatial analysis: where cultivated land concentrates across 775 Nigerian LGAs, with an explicit correction for uneven OpenStreetMap coverage — Moran's I = 0.300, the area effect surviving the coverage control (ΔAIC = 20). Spatial statistics on API-sourced data, frozen to CSV, 175 automated claim checks. Listed under featured projects above; manuscript in preparation for TeMA — Journal of Land Use, Mobility and Environment.
- Languages: Python, R, SQL, TypeScript/JavaScript
- Statistics and ML: PyMC, scikit-learn, XGBoost, TensorFlow/Keras, SciPy optimisation, survival, spatial and time-series models
- Apps and reporting: Streamlit, R Shiny, Gradio, React, Quarto, Tableau, Power BI
- Engineering: Git, GitHub Actions, Docker, Supabase, Cloudflare Pages, Render
♟️ Chess, peak 2000+: Chess.com rapid and Lichess blitz, as BayesianRules.



