BI Scientist building AI/ML systems · Lagos, Nigeria
precious.ikebude@outlook.com · WhatsApp +234 813 678 7181 · ikebude.github.io · Portfolio hub
I turn business data into dashboards people use, models that are measured honestly, and small services that run them. 3+ years of data, BI and finance-analytics work since Nov 2022; currently BI Scientist || AI/ML Engineer at 3Consulting Limited (Lagos), after starting there as a Junior Data Analyst in Jun 2025. B.Sc. Economics (2:1).
Open to: BI Analyst · Data Analyst · Data Scientist · junior ML Engineer.
Work: remote (Nigeria or worldwide) or Lagos; full-time roles and short contracts (dashboards, SQL and reporting, churn or risk models).
What I bring: Power BI and SQL reporting that answers a business question; models judged on out-of-time data against a simple baseline, with calibrated probabilities and stated limitations; and the engineering around them (FastAPI, MLflow, Airflow, Docker, CI).
| Project | What it is | Results (from the repo's reports/*.json) |
Stack |
|---|---|---|---|
| Moniguard | Fraud scoring service for pay-on-delivery and mobile-money transactions: a 0–1000 score with reason codes | On 60,000 synthetic transactions, time-based 70/30 split: PR-AUC 0.959 (95% CI 0.942–0.972) in batch evaluation and 0.837 when each transaction is scored on its own through the API, as in live use | Python, XGBoost, Isolation Forest, FastAPI, MLflow, Airflow, Docker |
| LoanLens | Credit default scorecard on public Prosper loans, with a leakage gate in tests and an approval-policy simulator | Trained on 41,638 earlier loans, tested on 13,880 later ones: ROC-AUC 0.705 (95% CI 0.695–0.714), KS 0.303 | Python, scikit-learn, Streamlit |
| BI Operations Agent | Ask business questions in plain English; answers come from a governed semantic layer through a read-only SQL guard, plus daily KPI anomaly checks | On synthetic retail data: 96.0% (48/50) in-sample and 56.0% (14/25) held-out questions correct, 12/12 unsafe prompts blocked, 36.6 ms p95 | Python, FastAPI, SQLite, sqlglot, Streamlit |
| Customer Intelligence | Telco churn model that ranks a weekly call list by revenue at risk, plus RFM segments | 7,043 customers (public data): PR-AUC 0.841, ROC-AUC 0.933, 80.5% precision when calling 20% of customers | Python, scikit-learn, Streamlit |
| AIOps Expense Tracker | NGN personal and small-business finance app, extended from a course starter | On synthetic data: categoriser macro-F1 0.920 (rules baseline 0.780), receipt-text field match 97.5% (text parsing, not OCR), forecast WAPE 8.1% (naive 11.2%) | Python, scikit-learn, FastAPI, React |
| Lolo's Ledger | Multi-asset trading journal and position-sizing risk engine with an AI analyst that never gives trading signals (source private; case study linked) | Qualitative: built on my desktop journal's engine and its tests | TypeScript, React, NestJS, PostgreSQL, Python |
More ML, SQL and Power BI projects, each with a short README and its numbers, are in the portfolio hub.
Shown as engineering case studies only; company source code is not public, and the public sites are deployment pending verification.
- FraudGuard (transaction fraud prevention): feature engineering for the scoring ensemble, Airflow and MLflow retraining, PostgreSQL-to-Kafka outbox, analyst dashboard. FraudGuard is the company's product; Moniguard above is my separate personal project.
- SafeData (NDPA/GDPR data retention and consent): retention-rule logic, PostgreSQL test-suite fixes, CI with linting, dependency auditing and secret scanning.
- Africa Stablecoin Intelligence (market and regulatory tracker): structured the market and regulatory indicators, Light/Dark/System theming, launch and indexing preparation.
- BI and analytics: Power BI (DAX, Power Query, star schemas, row-level security), Tableau, Excel (pivot tables, VBA), KPI definitions
- SQL: PostgreSQL, SQL Server, SQLite, DuckDB; CTEs and window functions
- Data science and ML: Python (pandas, NumPy, scikit-learn, XGBoost, statsmodels, SHAP), R; imbalanced classification, probability calibration, credit scorecards (WoE, KS), RFM and CLV, time-series forecasting
- ML engineering: FastAPI, MLflow, Airflow, Docker and Docker Compose, GitHub Actions; TypeScript and React for front ends
- B.Sc. Economics, University of Nigeria, Nsukka. Second Class Upper (2:1), Sep 2017 – Nov 2022
- Google Advanced Data Analytics Professional Certificate, Coursera, Apr 2025
- Managerial Economics & Business Analysis Specialization, Gies College of Business (UIUC), Coursera, Jun 2023
- Microsoft AI & ML Engineering Professional Certificate, Coursera: in progress, target Dec 2026
- Google Cybersecurity Professional Certificate, Coursera: in progress
- Planned next: Microsoft Azure and Google Cloud professional certificates (data science / AI engineer associate)
- Email: precious.ikebude@outlook.com
- WhatsApp: +234 813 678 7181
- Website: ikebude.github.io
- Data-science portfolio: datascienceportfol.io/ikebude
- GitHub: github.com/ikebude · portfolio hub ai-ml-data-engineering
- Kaggle: kaggle.com/ikebude