WORK
01
View case study
Password protectedDesign, Test, Enrich
Interaction design for a complex, multi-state flow · Self-initiated
Building an enrichment workflow meant configuring it blind across three screens and learning whether it worked only after a full catalog run. I proposed one loop instead, then streamlined it around a single rule: every project starts from a file, and every new file is a job.
02
View case study
Password protectedSelf-serve activation
Discovery from evidence · Journey design · AI behaviour as interaction
As Dyver opened up beyond ops-led enterprise delivery, new customers had to reach a first enriched catalog without a person guiding them. They stopped early, and not for the reason everyone assumed.
03
View case study
Password protectedCustomer journeys
Audience and funnel strategy · Design judgment beyond product UI
One homepage spoke to customers who need enrichment in two different ways, and the one who wants to start alone, today, had no clear way in.
04
View case study
Password protectedDesign system in code
Design-to-production · Design systems · AI-native practice
How one designer-engineer keeps a design system coherent when much of the code is generated: by enforcing it in the repository instead of checking against it afterwards.
05
View case study
Designing Adaptive Conversational Flows for an AI Health Coach
Conversational design · Behavioural science
Designing context-aware conversation patterns that adapt persuasion techniques to schedule stress
06
View case study
Multi-Agent Architecture for Intelligent Calendar Scheduling
Agentic AI architecture · Context engineering
Designing a 4-agent system from scratch: translating conversational flows into clean agent boundaries, typed delegation patterns, and state management for complex conversational AI
07
View case study
Interactive Data Tool for Property Developers
Product discovery · Data visualisation
Led product discovery for an interactive data visualization tool that helps property developers analyze location data, predict prices, and make informed investment decisions.
Agent behaviour
I write and fine-tune the prompts, tool contracts and conversational flows that govern how an agent asks, explains and recovers.
Interfaces in production
I build the interface those conversations happen in, directly in the production codebase. My handoff is working code plus the API changes a feature needs.
End-to-end product flows
From problem framing to shipped journeys, onboarding, activation and purchase, without a handoff in the middle.
Design systems in code
Tokens, theming, component architecture, WCAG and motion, enforced in the repo so new work builds on the system instead of drifting from it.
AI-native practice
Agentic coding tools every day, my own skills for repeating patterns, and code-generation recipes tied to the design system.
TOOLKIT
Design engineering
React, TypeScript, Next.js, Tailwind and shadcn, design tokens, component architecture, WCAG, motion and micro-interactions
AI and conversational design
Agentic and conversational flows, prompt design, agent harness, tool contracts, structured outputs, human-in-the-loop, behavioral science
AI-native workflow
Claude Code, Codex, Cursor, custom skills, code-generation recipes, Figma MCP, Git worktrees
UX and product design
Product discovery, user journeys, design systems, Figma variables and advanced prototyping, user research, data-driven iteration
Engineering and data
Python, FastAPI, PydanticAI, SQL, LangChain, OpenAI API, Docker