At OVO Energy, I consolidated three teams’ production services into an agentic monorepo and set up its agent harness and shared codebase. I also built an evaluation approach aligned to UK AI Security Institute methods.
I shipped the Customer Vulnerability classifier across every call, driving a 34% increase in Priority Services Register sign-ups and saving £4M+ per year. I also rescued an at-risk billing explainer agent and API, shipping it to a hard three-month deadline.
At CogCo. (The Cognition Company), I was tech lead for a GenAI research product and built its LLM-as-judge evaluation and Python prompt optimisation from scratch. This work secured an S&P 500 contract.
My PhD at the University of York focused on measuring player experience with eye tracking, ECG and psychometric compression. I bring that human measurement science into building and evaluating production AI.
