Problem framing & critical thinking
I find the flawed assumption early, when it's still cheap to fix. At Hitachi I co-organised a design sprint with data scientists and subject experts at Ofwat to identify workflow bottlenecks and prototype data-based solutions. At Sage I ran future-vision workshops with stakeholders and customers to shape a three-year product vision for the HR SaaS area.
AI prototyping & building
I build functioning, coded prototypes rather than mock-ups, so ideas can be tested with real users and real constraints. I've done this at Hitachi, UserZoom, Adestra and Versio4, for clients including Pearson, Chanel and UK government agencies, increasingly co-developing with AI tools such as Cursor, GitHub Copilot and Figma Make.
AI evaluation & trust
I design evaluations that answer whether an AI system actually works for the people relying on it: explainability, replicability, and fitness for real workflows. I bring statistical grounding (regression, clustering, correlation analysis, t-tests, A/B testing) and self-directed depth in how generative models work, including the mathematics of diffusion models.
Research & analysis
I use research as an input to building, not an end in itself. At Tesco, contextual observation, analytics review and iterative prototype testing of cash management software led to fewer financial errors and easier colleague training. At UserZoom I delivered quantitative research at scale for Shell, Bose, AXA and Commerzbank.
Responsible AI & governance
I help teams adopt AI responsibly: co-building tooling aligned to the UK government AI Playbook, and supporting colleagues through presentations, office hours and practical guidance. I've worked on services through GDS assessment and to WCAG AA.
