Coded and synthesised 40+ visitor interviews from the public evaluation of a heritage RAG virtual expert — finding that confident hallucination, not capability, was the real trust threat.
[ Evidence, not guesses ]
I research how people build trust with AI — then design the products that earn it. Research-led, evidence-first, shipped.
6 case studiesLondon, UK ( 51.5°N )Available for work
[ How I work ]
From question to evidence
Most design work starts with a solution. Mine starts with a question I can actually answer. Four stages, each producing something you can act on — not a deck, a decision.
01
Discover
- User interviews
- Field research
- Competitive teardown
- Trust & risk mapping
I start with your users and the thing you are actually unsure about. No pixels until the question is sharp.
02
Define
- Synthesis & coding
- Information architecture
- Personas & journeys
- Design brief
Interviews get coded, patterns get named. You leave with a written brief and an architecture, not a mood board.
03
Design
- UI & interaction design
- Design systems
- Prototypes in 48h
- WCAG accessibility
High-fidelity screens and a system that scales. Every decision traceable to something a user said.
04
Validate
- Usability testing
- Task success metrics
- Iteration rounds
- Handoff & QA
Tested with real people, measured against the task. If it does not move the number, it goes back.
[ Case studies ]
Selected work
An end-to-end AI agent that scans, diagnoses and autonomously fixes storefront issues — with configurable trust modes and a full audit trail.
A multi-client time tracking SaaS for a consulting firm, shipped and live. On-time submissions up 25%, QA errors down 20% in the first quarter.
A five-module analytics platform for small e-commerce owners who are not data analysts: dashboard, sales, analytics, inventory, settings.
Android app and web portal for a live sea port access system. Entry processing cut from 30 minutes to under 2. Over 10,000 Play Store downloads.
Led the research pipeline for a gamified cybersecurity escape room: a 29-question survey, binary coding of 98 responses, five in-depth interviews, and personas that shaped the game.
78%skip password updates
51.5%no VPN on public Wi-Fi
62%rate those same threats top risk
[ Outcomes ]
Numbers, not adjectives
Measured results from shipped and evaluated work. Where a figure comes from a concept study rather than production, the case study says so.
| Task success rate at concept stage | 90% |
|---|---|
| Increase in user engagement, time tracking platform | +25% |
| Reduction in QA errors, first quarter | −20% |
| Port access processing time | 30 → 2 min |
| Play Store downloads, shipped app | 10,000+ |
| Visitor interviews coded, REX evaluation | 40+ |
| Survey responses binary-coded | 98 |
| Field research on human trust in AI | 6 months |
[ Research background ]
Where others guess, I have evidence.
I spent time in academia studying the exact problems AI product teams are now shipping into. That background is not a detour — it is the differentiator.
Most designers cannot walk into an AI startup and talk about how users build trust with a model that sometimes invents things. I can, because I studied it in the field.
- ( R1 )Trust in AI chatbots using LLMs for museum visitors
- ( R2 )Hyperlocal AI chatbot — walking companion app
- ( R3 )Smart cities — accessible public services
[ About ]
Detour through research. Back with evidence.
I am a product designer with a research background in AI trust and human-computer interaction. I have shipped across SaaS, e-commerce, smart cities and accessibility — always with measurable outcomes.
I use AI tools to move fast, but the judgment about what to build and why is still mine.
- Figma
- Framer
- Prototyping
- Design systems
- WCAG
- Material Design
- User interviews
- Usability testing
- Dovetail
- Affinity mapping
- AI synthesis
- Field research
- v0.dev
- Bolt
- Claude
- Lovable
- Figma AI
- Otter.ai
[ Contact ]
Let's build something worth testing.
Open for freelance product design, AI product consulting and UX research sprints.