Ten years designing for product, three spent deep in AI — RAG, GenAI, and semantic search for UK government, defence, and enterprise. My job is to take something powerful but opaque and make its reasoning legible: what it found, how sure it is, and where the answer came from.
Each one had to earn trust before it earned use. The thread running through them: surface the evidence, score the confidence, and never hide uncertainty behind a single confident answer.
FCDO — Foreign, Commonwealth & Development Office · via Faculty AI
A foreign-language propaganda monitor that pairs RAG and keyword search with auto-translation and summarisation — so non-Arabic-speaking analysts read answers in minutes, not days.
FCDO — Foreign, Commonwealth & Development Office · via Faculty AI
A conversational platform that lets analysts question huge, sensitive datasets in plain language — returning instant summaries, network maps, and explainable results, each traceable to its source.
DSTL — Defence Science & Technology Laboratory · via Faculty AI
One platform for DSTL's whole innovation pipeline — from spotting novel research to running structured experiments. AI shortlists the 10 best-fit papers from thousands; it became the primary system within 15 months.
FCDO — Foreign, Commonwealth & Development Office · via Faculty AI
A validation platform that scores AI translation confidence and routes the uncertain cases to expert linguists — replacing scattered manual handoffs with one shared, accountable workflow.
I design for the moment a person has to decide whether to believe a machine — and I think that decision should never be a leap of faith.
Most of my work lives where the stakes are real: intelligence, defence, government accountability. In those rooms an AI answer isn't useful unless the person can see why it's there — the evidence behind it, how confident the system is, and what it's unsure about.
So I design the seams, not just the surface. Confidence scores, source traceability, layered disclosure, human-in-the-loop checkpoints. The aim is never to replace expertise; it's to give experts a faster, clearer way to apply it.
I work end-to-end — research and journey mapping through to high-fidelity UI in the GOV.UK and Carbon design systems. I'm freshly back from seven months in South America and looking for the next hard problem worth solving.
Turned an offline retailer-dispute service into a fully self-serve digital journey, rebuilding the information architecture around what customers were actually trying to do.
A QR-triggered landing page and a non-intrusive signup prompt for Dove's new brand. The winning variant lifted signup rate by 20%.
Redesigned end-to-end booking and amend flows across web for leisure and business guests, cutting friction at the points where bookings were being lost.
Led discovery across complex operational workflows, then designed a future-state experience to digitise and streamline internal service delivery.
Whether you're building an AI product, untangling a complex government brief, or just want users to feel surer about what your product is telling them — I'd love to hear about it.
