Backonaut – Lightweight Backtesting Engine

Designed extensible OOP components for Strategy, DataFeed, Portfolio, and Execution. Integrated Yahoo Finance API for live historical data. Built custom strategy classes and automated PDF reporting.

GitHub

Monocular Depth Estimation via Transfer Learning

Developed three depth estimation approaches as an alternative to expensive LIDAR (saving car owners $1000+). Compared Pix2pix, U-Net with DenseNet, and U-Net with MobileNetV2.

GitHub

Food & Nutrition Analytics

Analysed 587K days of MyFitnessPal data across 9,900 users. Derived insights to aid supermarkets in stocking — potentially reducing 10,000+ tonnes of food waste.

GitHub

Fraud Detection using Graph Machine Learning

End-to-end Graph ML solution to detect fraud on a large financial dataset by Vesta. Leveraged graph-based representations for transaction network analysis.

GitHub

Query Optimizer

Built a query optimiser and estimator for SJDB (designed by Dr Nicholas Gibbins). Implemented cost-based optimisation strategies for relational queries.

GitHub

Tweet Classification – MediaEval Benchmarking

Theorised and critically analysed five NLP classification solutions for real/fake tweet detection on the MediaEval 2015 dataset.

GitHub