I build the intelligence layer between raw data and executive decisions.
Seven years. Four industries. Three continents. From public health systems in Nigeria to fintech ops in the US, I architect AI-powered analytics that don't just report what happened but shape what happens next.
Most data professionals hand you a dashboard and call it done. I build the systems behind the systems.
Multi-agent AI pipelines, product intelligence frameworks, and operational analytics architectures that give leadership real clarity, not just real-time numbers.
With a B.Sc. in Physics from FUTA and an active MSc in Data Science, I bring rigorous analytical thinking to every layer of the stack, from ETL pipelines and SQL models to LLM prompt architecture and cross-functional performance strategy.
I've embedded intelligence into public health operations in Nigeria, led product analytics at scale in US fintech, built branded analytics programmes in the UK, and shipped open-source AI systems that benchmarked to production standards.
Currently operating at the intersection of product, data, and AI and open to senior roles across the US and EMEA.
Built a production-grade multi-agent AI system to coordinate national emergency dispatch across 8 Nigerian cities, handling 89 concurrent incidents, 3 operational shifts, and real-time cross-city resource conflict detection.
The architecture deliberately assigns specialised agents to distinct tasks rather than overloading a single model, proving that role-based agent orchestration outperforms monolithic LLM approaches on complex, time-critical operational problems at scale. Fully open-sourced for the AI and emergency management community.
Each agent handles a distinct operational role - dispatch coordination, resource allocation, conflict detection, and shift handover - with no single model overloaded. Deployed and validated in a containerised environment against a structured multi-city incident corpus.
Built a production-grade fraud detection and risk scoring system aligned with CBN, Visa, and Mastercard compliance frameworks. Combines an ensemble ML layer (XGBoost, Random Forest, Logistic Regression) with a deterministic rules engine and a 3-zone blend scoring model calibrated against four real-world fraud datasets.
Deployed as an interactive Streamlit application with single-transaction scoring, bulk upload analysis, KRI dashboards, and a Monte Carlo stress-testing module with configurable rule slip rates and VaR output.
Ensemble weights calibrated across PaySim, Credit Card, BAF, and Nigerian fintech data. Rules engine and ML scores blend across three risk zones - GREEN, AMBER, RED - with Monte Carlo VaR stress testing at configurable slip rates.
Analysed customer retention trends for an e-commerce business by segmenting ~500,000 UCI Online Retail transactions into monthly cohorts. Identified drop-off points, systemic retention failures, and the best-performing acquisition periods to guide lifecycle strategy.
Analysed 1,338 customer records to identify the key drivers of insurance premium costs. Built a data story and agent guide using regression, correlation, and univariate analysis, answering management questions on gender, region, smoking, age, and BMI impact on rates.
Built an end-to-end analytics pipeline for Toman Bike Share, consolidating two years of ride data via SQL, enriching with cost tables, and delivering a Power BI dashboard covering hourly revenue, seasonal performance, profit margins, and rider segmentation. Concluded with a data-backed pricing strategy recommendation.
Cloud-based and local server-based solution for hospitals, clinics, pharmacies, and laboratories. Contributed to data architecture, analytics, and system design at MTS Digitals.
A deep-dive into why monolithic LLMs break under real operational pressure - and how role-based agent orchestration solves problems that single models can't.
Accuracy is a metric. Usefulness is a decision. This piece unpacks when optimising for the wrong objective destroys value - and how to think differently about model evaluation.
Drawing on a background in Physics to reframe how data behaves - patterns, forces, and the laws that govern how information moves through systems.
A practitioner's notes on where databases end and caches begin - why the boundary matters more than either layer, and how to reason about it in real systems.