Senior Data Engineer & Scientist | Building data platforms, ML systems, and AI agents that make it to production.
π Lagos, Nigeria Β· π Open to remote contracts (EU / UK / Canada/ Global) π« erozonkwo@gmail.com Β· LinkedIn Β· GitHub
I've spent 10+ years building the systems that quietly power fintech, banking, Energy, FMCG and telecoms such as fraud detection engines, credit scoring models, next best action/recommendations engines, chatbots, AI Agents and pipelines that process billions of transactions. I work across the full data lifecycle: from raw ingestion and cleaning, to production ML, to the dashboards executives actually use.
Lately I've gone deep into LLMs and agentic AI β building workflows where models don't just answer questions, they get things done.
I care about reliable, well-tested, well-documented data. I don't over-engineer, I ship, and I write code thinking about whoever inherits it.
Data Engineering
Apache Spark / PySpark Β· Kafka Β· Hadoop / HDFS Β· dbt Β· Airflow Β· Snowflake Β· BigQuery Β· ELK Stack Β· SQL / NoSQL Β· ETL/ELT Design
Cloud & Infra
GCP Β· AWS Β· Azure Β· k8s. Databricks Β· Docker Β· Kubernetes Β· CI/CD (GitHub Actions) Β· Git Β· REST APIs
Machine Learning
Python Β· PyTorch Β· TensorFlow Β· scikit-learn Β· XGBoost Β· SHAP / XAI Β· Clustering (DBSCAN, K-Means) Β· NLP
AI / GenAI
LLMs Β· Autonomous AI Agents Β· RAG Β· Prompt Engineering Β· Recommender Systems ,Langchain,n8n. etc.
MLOps
MLflow Β· Model Monitoring & Drift Detection Β· Feature Stores Β· Reproducible Pipelines
Analytics & Viz
Power BI Β· Looker Β· Tableau
- β‘ Real-time fraud detection platform β PySpark β Kafka β Airflow pipeline processing 50M+ transactions/day at sub-second latency
- π€ AI-as-a-Service platform β containerised model-serving layer exposing risk scoring, churn, and merchant analytics APIs to fintech partners
- π ELK-based log analytics platform β near real-time network fault detection across 30M+ subscribers
- π Automated settlement pipeline β replaced a 58-day manual process with a 2-hour automated run at 99% lower error rate
- π― Customer analytics suite β sentiment analytics, churn, propensity, RFM, and CLV models driving 29% churn reduction and 57% loan sales growth
- π Group-wide industrial analytics platform β unified data models spanning cement, fertilizer, energy, and petrochemicals on a centralized warehouse
- π± Digital analytics engine β AI-driven loan scoring, churn prevention, recommendations, and CLV management β 22% growth in digital product acquisition
- π Automated incentive settlement pipeline (Python + SQL) β replaced a 28-day manual process with a 2-hour automated run, 99% error reduction
- π MLOps standardization β accelerated model-to-production cycle by 70% via standardized pipelines, feature stores, and CI/CD enforcement
- π XAI/governance framework (SHAP-based) β cut model review & regulatory approval timelines by 70%, sustained 99.9% model uptime via drift detection and automated retraining
- π― Customer analytics suite β churn, propensity, RFM, telemarketing optimization β 18% churn reduction, 37% loan sales growth, 26% uplift in campaign response, conversion improved 21% β 59%
- π§© Customer segmentation (K-Means, DBSCAN) β powered 13 new tailored business propositions across sales, marketing, and CX
- π΅οΈ Facial-matching fraud detection (DBSCAN on KYC embeddings) β eliminated duplicate SIM registrations, hardened identity verification at scale
- π³ Loan pre-scoring model validation β built XGBoost challenger models achieving +8pp KS improvement; applied SHAP-based XAI for regulatory/credit-committee confidence
β οΈ Quantitative risk modeling β VaR, Monte Carlo simulation, stress testing supporting ICAAP and Basel II/III capital submissions
| Metric | Result |
|---|---|
| Model deployment speed | 70% faster |
| Pipeline error reduction | 99% |
| Production model uptime | 99.9% |
| ETL turnaround | 28 days β 2 hours |
| Conversion uplift | 31% β 59% |
- Building agentic AI workflows that automate real business tasks
- Scaling dbt-based ELT architectures
- Exploring how LLMs can augment traditional analytics , not replace it
ποΈ Queen's Young Leaders Award (2015) Β· Associate Fellow, Royal Commonwealth Society (2016) Β· YALI West Africa (2017) Β· CEO Recognition Award (2024)
π¬ Got a hard data problem and not enough hands? Let's talk.

