Student Data Researcher

The brief

A disaster-management research group had data from five sources that disagreed with each other about almost everything: units, column names, date formats, what counted as a missing value.

I built the Python workflows that reconciled them, then the models on top. The interesting work was upstream — by the time the data was clean, the modelling was comparatively ordinary.

What I did

• Architected and deployed Python workflows that automated cleansing and integration across 5+ multi-source datasets.
• Lifted validation accuracy 25% over the first working baseline through feature engineering and hyperparameter tuning, on models built over Pandas, SciPy and SQL.
• Refined reporting with complex SQL — CTEs and views — surfacing results through Power BI and Tableau for decision-makers.

Measured

Architecture

In hindsight

I validated the cleaned data by eye and by spot-check. With hindsight I'd have written the expectations down as tests — row counts, value ranges, join cardinality — so that a bad upstream refresh failed loudly instead of quietly shifting the model.

Stack