Multimodal research and engineering in Gemini, with a particular focus on vision encoders and multimodal understanding.

Worked in the ML Tech Team where cutting edge developments in machine learning and artificial intelligence were explored for optimised edge inference on ARM ML chips.

A summer research placement in the Biomedical Image Computing subsection of Computer Vision Lab. I investigated the use of recursion in training convolutional neural networks for semantic segmentation of short-axis cardiac MRI using weakly annotated data, with the end goal of improving data efficiency of ML based tools.Using data that took less than 10% of the time to annotate, I was able to achieve results which are 96% as accurate as those achieved by the fully supervised counterpart, ultimately leading to "Learning to segment medical images with scribble-supervision alone" which was accepted in and received the Best Impact Award at DLMIA2018.

Designed & built a Microsoft Access based VBA/SQL database/software package for management of complex chemical 'recipes' for protein crystallization screens, automating production of vital safety, preparation and sales documents.Excerpt from reference:"Basil produced for us a beautiful piece of software that fulfilled all our needs and much more. It is amazing to use and has changed our way of working for the better. Basil was hands on all the way through the project and worked closely with different people from different departments. He was exceptionally on-the-ball and proactive and totally un-phased by anything thrown at him."

A 13 month pre-university placement undertaken with DNV GL's Risk Advisory department in Loughborough. Assisted with quantitative risk assessments on a variety of oil & gas assets, from LNG terminals in Angola to mainland gas transmission and distribution pipelines. Fairly extensive use of VBA with Excel/Word in order to create both internal solutions to speed up risk assessment procedures and products for client distribution.