Senior Data Analyst | Data & AI Engineer | Evidence-Traceable BI, Machine Learning & Software Engineering
Data x AI x Software
Doctoral student · M.Sc. Data Science (Northeastern) · M.Sc. IT · B.Sc. Accounting
I build systems that prove themselves — every number traces to a source, every decision has a record.
(Click to test the APIs & Play the Game)
- Payments API
- Fraud Detection Engine
- Reconciliation LLM Layer
- Neon Drift 3D Game
- Credit Risk Dashboard (React UI)
A system is only as good as its ability to survive uncertainty.
Whether I am architecting a double-entry ledger, forecasting credit risk, or building a 3D game engine, my philosophy remains the same: Traceability before intelligence. Explain a number you can already prove, not one you're hoping is right.
My career has been a deliberate progression across three disciplines since 2018:
- Accounting (B.Sc. & Deloitte): Taught me that every number must have a source and a counterparty.
- Data Science (M.Sc. & Wizeline): Taught me how to find patterns in those numbers.
- Software Engineering (Primero & Contract): Taught me how to turn those patterns into systems that run in production.
Today, I operate at the intersection of Data Science, AI Engineering, and Software Engineering, building decision engines for fintech, payments, and logistics.
| Project | Business Impact | Stack |
|---|---|---|
| credit-risk-forecasting-engine | The Business Question: "How much money will we lose next quarter, and what is our risk tolerance?" Built a system that calculates Expected Loss (PD × LGD) with 90% confidence intervals. Uses rolling-origin backtesting and asymmetric cost functions to output a "Safe Capital Reserve." Features a Rust inference engine for sub-10ms latency. |
SQL, Python, XGBoost, Rust, statsmodels |
| credit-risk-dashboard | Executive React frontend for the Credit Risk Engine. Visualizes Expected Loss forecasts with 90% confidence intervals via Recharts. | React, Vite, Recharts, Vercel |
| Project | Business Impact | Stack |
|---|---|---|
| Enterprise-RAG-pipeline | Production-grade RAG system with hybrid search, access-control-aware retrieval, and prompt-injection defenses. Features a custom evaluation harness and canary deployments to prevent stale data. | Python, Vector DB, LLMs, FastAPI |
| reconciliation-llm-layer | Traceability-constrained LLM narration. Uses Pydantic to strictly limit the AI to only the numbers provided in the input, eliminating hallucinations in financial reporting. | Python, FastAPI, Gemini |
| Project | Business Impact | Stack |
|---|---|---|
| principal-sql-analytics-engine | Production-grade SQL engine using advanced window functions, recursive CTEs, and cohort logic to model customer lifetime value, retention, and unit economics. | SQL, DuckDB, Python |
| data-analytics-portfolio | Flagship Repo: 6 end-to-end analytics projects covering SQL modeling, executive financial dashboards (Amazon, JPMorgan, Walmart), healthcare metrics, and retail profitability. | SQLite, Python, Excel, Power BI |
| Executive-Sales-Dashboard- | Executive sales dashboard with KPI traceability, pivot-table evidence, and regional performance analysis. | Excel, pivot analysis |
| Toyota-FY2026-Executive-Disclosure-Dashboard | Source-linked financial analysis and executive dashboard with KPI traceability and SEC filing citations. | Excel, Public Filings |
| operational-bottleneck-system | Logistics bottleneck intelligence using throughput, queue dynamics, and financial impact scoring. | Python, Excel |
| forensic-healthcare-billing-screen | Peer-group anomaly screening for healthcare billing. Hierarchical peer fallback, robust z-scores, composite screening score. 18 passing tests. | Python, pandas, Excel |
| Project | Business Impact | Stack |
|---|---|---|
| frontierswe-rate-limiter | Principal-level AI evaluation task requiring a high-concurrency distributed rate limiter. Proves race condition mitigation (20 concurrent requests bypassing a limit of 5) and verifies atomic fixes. | Rust, Python, FastAPI |
| payment-api | Production-grade, event-sourced ledger with double-entry accounting and immutable audit trails. Handles idempotent payments, state machines, and HMAC-verified webhooks. | Python, FastAPI, SQLAlchemy, PostgreSQL |
| reconciliation-engine | O(n) transaction reconciliation engine. Hash-map matching, five-status classification, and value-at-risk reporting for high-volume settlement files. | Python, FastAPI, pytest |
| fraud-detection-engine | Real-time API scoring transactions with calibrated thresholds. Built an explainability layer (SHAP) ensuring every decision can be audited and defended to compliance teams. | Python, scikit-learn, FastAPI |
| lstm-fraud-detection | Experimental sequence-based deep learning model for fraud detection. Analyzes the temporal sequence of a user's last 10 transactions to spot patterns over time. | Python, TensorFlow, Keras |
| Project | Business Impact | Stack |
|---|---|---|
| neon-drift-game | High-performance 3D browser game built from scratch. Features real-time collision detection, particle physics, object pooling for memory management, and a custom delta-time game loop. | JavaScript, Three.js, WebGL, Vercel |
| credit-risk-dashboard | Executive React frontend for the Credit Risk Engine. Visualizes Expected Loss forecasts with 90% confidence intervals via Recharts. Live Demo | React, Vite, Recharts, Vercel |
Core Engineering
Data Science, ML & AI
Infrastructure & MLOps
- Senior Data Analyst (Contract): Building SQL pipelines and automated dashboards for executive stakeholders. Ensuring evidence-traceable reporting.
- Full Stack / Data Engineer @ Primero: Architected event-sourced payment systems and O(n) reconciliation engines. Managed CI/CD pipelines.
- Senior Data Scientist @ Wizeline: Built end-to-end forecasting systems with rolling-origin backtesting.
- Data Analyst @ KPMG: Performed financial modeling and anomaly detection for enterprise audit clients.
- Financial Data Analyst @ Deloitte (2018): Analyzed large financial datasets to identify discrepancies, cost-saving opportunities, and internal control weaknesses.
(For the full interactive timeline, see my LinkedIn)
The interesting engineering is not in the happy path. It is in the failure modes — the concurrent request, the invalid state transition, the empty input, the forged webhook. Systems that prove themselves are systems that handle what goes wrong.
Open to remote-first financial engineering, Data Science, Machine Learning, and analytics roles. Available immediately.
Lagos, Nigeria (WAT/UTC+1). Building fintech and logistics systems since 2018. I turn messy operational data into auditable pipelines, transparent models, and decisions someone can defend.