Ekundayo Onifade

Ekundayo Onifade

Analytics, applied machine learning and data engineering

London

I work on messy real data: cleaning it, modelling it, and writing up what the numbers actually support rather than what they seem to say. MSc Data Science and Analytics, University of Westminster.

Work

Customer analytics: segmentation and retention

1.07 million transactions, UK wholesale, 2009 to 2011

Retention heatmap showing 24 monthly customer cohorts

I expected retention to fall away steadily after acquisition. It did not. Most cohorts retained better in month two or three than in month one, because retailers reorder on their own stock cycle.

A standard 30 day churn rule would have flagged healthy accounts as lost. The analysis also found the source table was not a transactions table at all, but an operational log holding four different record types under one schema.

Read the full analysis

UK property analytics pipeline

An end to end pipeline over HM Land Registry data in dbt Core and DuckDB, with automated data quality tests and documented lineage. Includes a Tableau choropleth and a Gemini API reporting layer that verifies every generated figure against source before publication.

Chess anomaly detection

Unsupervised anomaly detection across 48,933 games using Isolation Forest and K-Means, with thresholds derived rather than assumed. MSc thesis.

House price prediction

Regression models with feature engineering and hyperparameter tuning to forecast residential property prices.