About

For my masters thesis, I developed deep learning models to optimise bike-sharing usage predictions across London’s Santander Cycles network. This was the first application of such models on this dataset, including novel architectures like hybrid graph convolutional long short-term memory networks (GC-LSTMs). By preprocessing over 60 million trip records (massive pain given the low quality data), I built and benchmarked models that outperformed traditional methods, achieving high prediction accuracy which can improve redistribution efficiency. Sadly it was not adopted due to big organisation / bureaucratic issues – so still no bikes available when I need them but I learnt lots from the project (considering I started with effectively no coding experience). 

I worked on the foils and control team for a human powered hydrofoil project. I designed a novel front foil system (responsible for pitch stability and lateral control), decided the material design and personally manufactured the foils (using a 3D printed core wrapped in carbon fibre). We managed to create a carbon fibre system that was orders of magnitude lighter than the previous design whilst creating function control systems and reducing overall drag.

I was introduced to the Global Himalyan Expedition through a friend and the founder wanted to explore the opportunity to partner with a major drinks company. I was able to design and build a prototype solar water heater using recycled drinks bottles and locally available cheap materials to replace the expensive heating systems they were installing in remote villages in northern India.