CPA Finalist (Kenya) turned data scientist — 3 years inside development-finance operations, now building the credit-risk and fraud models that sit downstream of that work.
I spent three years at GIZ managing receivables, reconciliations, and MI reporting for a development-finance portfolio — the kind of work that teaches you exactly where financial controls break and what a risk team actually needs from a model. I'm now building that side too: credit scorecards, fraud classifiers, and portfolio analytics in Python, validated the way an auditor would validate them.
| Resource | Why look here |
|---|---|
| Fraud Detection System | The project to actually judge me on. XGBoost fraud classifier on 6.3M PaySim mobile-money transactions, with a training pipeline that literally could not run until I found and fixed the bug — documented in the commit history, not hidden. |
| Financial-Analyst | Three-statement models and DCF valuations built from primary-source SEC filings, with a validation tab tying every historical line back to source. |
| Stock-Portfolio-Tracker-Analytics-Engine | CFA-level risk analytics in Excel — VaR (parametric/historical/Monte Carlo), CVaR, Black-Litterman optimisation, a 23-test validation suite. |
| Full work history: GIZ finance operations, CPA, McKinsey Forward. |
Current role: Finance Specialist, GIZ (Deutsche Gesellschaft für Internationale Zusammenarbeit) — Nairobi · 3 years in development-finance operations: AR/receivables management, reconciliation, and MI reporting dashboards built in SQL and Excel for a development-finance portfolio.
Credentials: CPA Finalist, Kenya · BSc Finance (The Co-operative University of Kenya) · Data Science, ExploreAI Academy (completed March 2026) · McKinsey Forward Program.
The combination I bring isn't "another data scientist" — it's someone who has sat inside the finance-operations function a credit or fraud model actually has to serve, and who can now build that model too.
| Project | What it does | Status |
|---|---|---|
| Fraud Detection System | XGBoost fraud classifier on 6.3M PaySim mobile-money transactions — balance-discrepancy feature engineering, isotonic calibration on a held-out split, 100% precision / 88.6% recall at the deployed operating threshold | Public |
| Financial-Analyst | Company financial models built from primary-source filings — three-statement models, DCF valuation, scenario toggles, source-linked validation tabs | Public |
| Stock-Portfolio-Tracker-Analytics-Engine | Portfolio risk/performance analytics engine in Excel 365 — CAPM, VaR/CVaR, Sharpe/Sortino/Calmar, Black-Litterman, tax-aware rebalancing | Public |
| Credit risk & alternative scoring portfolio | WoE/IV scorecards, SACCO stress testing, and DFI-focused credit modelling — the applied side of the CPA + data-science combination | Building next |
Credit risk: WoE/IV, scorecard development, GINI/KS/PSI, IFRS 9 ECL. Fraud: imbalanced classification, cost-sensitive thresholding, PR-AUC-first evaluation. Finance: GAAP/IFRS, 3-statement modelling, DCF valuation.
Open to Credit Risk Analyst, Data Analyst, and Financial Data Scientist roles — particularly where the finance-domain depth and the modelling capability both matter.
