Clyde
Portfolio agentWelcome, what would you like to know about Andy's work?
Checking Clyde...
The static page is available, but the Clyde API is not reachable from this view. Run npm run dev locally or deploy the API functions to unlock the prompt.
Scroll down to find out more...
AI product design specialist
AI product design, with the evidence left in.
I design AI-enabled products and workflows where the hard part is not the prompt, but the system around it. The fundamentals have not changed: source quality, interaction design, guardrails, evaluation loops, and workflow fit still decide whether the product works. AI is where that complexity is moving, so that is the layer I design and build with today.
- AI productRAG assistants, agent workflows, prompt surfaces, and design-to-code systems
- Evidence-firstResearch, analytics, prototype artefacts, evaluations, and source citations
- Safe by designGuardrails, trusted sources, evaluation, and human review built into the workflow
AI product surface
The portfolio is a working assistant, not just a set of pages.
Input
curated Markdown sources, metadata, confidence, and public citation URLs
Model
retrieved context is sent to Gemini server-side with answer and refusal rules
Output
answers return with citations, source notes, rate limits, and no raw PDF links
Critic system
AI-product energy, filtered through product-design judgement.
The brand borrows from AI tools such as Stitch, v0, and Claude: prompt surfaces, canvas motion, agent state, fast iteration, and visible system context. The critic keeps it bespoke, readable, accessible, and grounded in evidence.
AI interaction
Prompt-first flows, suggested questions, citations, model status, and useful empty states.
Brand and copy
Techy and direct, with concrete AI product nouns instead of vague transformation language.
Motion polish
Kinetic grid, signal sweeps, active surfaces, and reduced-motion fallbacks.
Source boundary
Reviewed public sources, privacy refusals, server-side model calls, and no raw PDFs in public HTML.
Selected AI outcomes
Recent work turning AI into inspectable product systems.
These projects are not isolated prompts or prototype screens. They are operating systems for AI product design: reusable evidence, agent workflows, evaluation loops, and artefacts that other teams can inspect.
£805M
annual MedExpress UK funnel represented in the design and analytics work
9
weight-loss GLP-1 personas built with population share and cited evidence
22
UK healthcare competitors modelled in the crawler and pricing intelligence surface
41
iterations of the shared MedExpress design-agent bundle shipped to the team
Case studies
Two HeliosX case studies in AI product design.
AI product designer / system architect / HeliosX
Designing Leo and the AI-assisted acquisition workflow for MedExpress UK
I rebuilt acquisition design practice around a paired knowledge base, RAG-ready source model, generated Claude Project bundle, and prototype quality gate: analytics, customer research, clinical gates, brand rules, and AI critics feeding the same reusable design memory.
- Closed nine open analytics questions across the new-customer journey.
- Built a 60-component rebrand recreation as a readable design-system source of truth.
- Created a multi-agent quality gate with generator, coordinator, specialist critics, and deterministic checks.
AI product designer / Gemini + Playwright system
Competitor Intelligence for regulated medical-intake flows
A TypeScript, Playwright, and Gemini intelligence engine that walks healthcare competitor journeys as PASS and FAIL personas, caches answer maps by page hash, asks the model only when a form is novel, and alerts the team when a competitor's flow materially changes.
- Modelled 16 patient personas across treatment verticals and clinical gates.
- Persisted page states, route edges, consultation paths, and change events to Postgres.
- Generated a 22-brand UK GLP-1 pricing analysis from repeatable evidence capture.
Practice
What I bring into product teams.
I am most useful where AI needs to become a real product capability: regulated workflows, fragmented data, ambiguous source quality, customer research that needs synthesis, and teams trying to move fast without losing accountability.
01
AI interaction design
I design prompt surfaces, answer contracts, citations, model status, review loops, and escalation paths.
02
Evidence architecture
I turn research, analytics, compliance, and stakeholder decisions into source systems that AI and teams can reuse.
03
Agentic production
I build workflows where agents help with synthesis, prototypes, audits, and reviews while deterministic gates keep output accountable.
04
Regulated product craft
I design AI-enabled workflows for commercial outcomes while respecting clinical safeguarding, advertising rules, accessibility, and trust.
About
An AI product designer comfortable with ambiguity, systems, and the last mile.
I design AI-enabled products by connecting what customers need, what the business is trying to change, and what the system can safely support. My recent work has focused on MedExpress UK and HeliosX, where the work spans acquisition, consultation, checkout, clinical review, retention, design-system memory, RAG assistants, and competitor intelligence.
The thread through the work is simple: make the important context easier to find, make the model output easier to inspect, and make the next version start from a higher floor than the last one.