At dunnhumby, I analyzed Tesco sales and shopper behavior data, uncovering an underperforming chilled-desserts product and delivering strategic recommendations for major UK retail clients. I also contributed to promotion optimization across 20+ historical promotions by identifying data gaps for promotional modeling.
I've built predictive models in Python and SQL, including a Random Forest fantasy-points model that reduced mean absolute error to 4.85 using external market signals. In my computational science practicum, I trained a multi-task CNN on 1.36M brain MRI images, reaching roughly 99.5% imaging-plane classification accuracy and a fourfold gain in localization precision.
