Building foundation models for galaxy images. Data-centric approach via crowdsourcing at scale (100M+ volunteer labels) and SSL.

Lead data scientist at Galaxy Zoo building and deploying ML systems. Created ML models for Euclid space telescope pipeline (July '23 launch). Led grants from European Space Agency, Meta, Alan Turing Inst ($180k total).

Secondment to Private Office of the Secretary of State, supporting ministers in making urgent decisions on the NHS emergency healthcare system. Funded and recruited by 10 Downing Street Data Science ("10DS").

Working at fintech startup Cytora, I used messy open data and machine learning to optimise commercial property insurance pricing. I created a strategy to increase the profit of a client insurer by circa £1.3m. Cytora is now expanding from 25 to ~80+ employees and raised a new investment round.

Edinburgh Airport has detailed records of flight activity, (both via Edinburgh and worldwide), ticket pricing, aviation and tourism market trends, and so forth. My role was to use this dataset to create highly quantitative business cases to present to airlines. I also developed software tools to help my business-trained colleagues do the same.On my own time and initiative, I created data acquisition software (Python) with competitive functionality to a €13,000pa subscription software. Please note that the specific details of this work are commercially sensitive and subject to a non-disclosure agreement. My internship was organised through the Saltire Foundation, a not-for-profit aimed at supporting outstanding undergraduates to become “the next generation of Scottish leaders”.

Extended data analysis software package of leading dark matter detector LUX (South Dakota, USA). Added capability to detect dark matter with more complex interactions than other experiments regularly search for. Implemented Effective Field Theory interaction form factor description in C++/ROOT analysis code. Supervised by Dr Paolo Beltrame. Funded by Royal Astronomical Society and University of Edinburgh.