Responsible for improving the accuracy of ANZ’s property price prediction models. Including feature engineering (Python/Pandas), building data pipelines (Java/Spring), and model serving (AWS Lambda/Node.JS).
•
Built Java Spring Boot data pipelines, which more than doubled the available training data for the property price prediction models. The impact of having more training data was a significant increase in both prediction accuracy and number of predictions possible thanks to an increase in overall real estate market coverage.
•
Developed the React theming system for REALas.com allowing for the appearance of the website to be changed dynamically for marketing campaigns and seasonal promotions. This replaced the original system which involved very large manual deployments and allowed for the theme of the entire site to be toggled via a single parameter.
Maintained and developed features for a large legacy PHP / MySQL internal tool. I was responsible for both building new MVC features as well as maintaining and patched the on-premises Ubuntu servers on which this service was hosted.
•
Optimised SQL queries running through a legacy ORM to improve query speed by almost 80%.
