Case studies
Rakuten Advertising's affiliate tools were disjointed, form-based SaaS dashboards. Mirai replaces them with a team of specialised agents and a portable design system that create, manage and guide brands’ affiliate programs through conversation and HITL.
Disjointed SaaS dashboards built around static, form-based tasks and disparate data. The real work wasn't in creating a chatbot, it was rethinking the tasks themselves, using AI innovations to stay ahead of the industry.
UX Design Manager, leading vision and design direction with an SVP sponsor, Product Management, Engineering and QA. There was also close collaboration with Data Science on the MCP endpoints behind each agent.
User interviews and data defined the tasks we prioritised automating. Agent endpoints were validated with Data Science before design. A portable, MCP-based design system was built alongside. This design approach became the test-bed for every AI application the business built moving forward.
Before any screens, I mapped how a conversation with an agent should unfold, what it asks, the choices a person makes and where an edit loops back. Early flows like these kept design, product and engineering aligned on the whole exchange, also surfacing the importance of retaining human-in-the-loop (HITL) rather than pure conversation.
A key design question was what an agent's answer looks like when words aren't enough (or as we frequently discovered, too much). Alongside text, agents can return small interactive cards inside the conversation. These could be a performance chart, action buttons, pre-defined responses. We discovered that this HITL approach of acting on an answer, not just reading it, often sped up interactions and increased a sense of trust.
Some answers are clearer as a chart or a preview than as a paragraph.
Built on an open standard, so a card works wherever the conversation happens.
Conversation helps with the task in hand. A dashboard you keep open to monitor your programme is a separate surface.
Early steps in the initiative contributed to a $200M increase in network gross sales. Creating an offer is now 93% faster than it was in the legacy form-based dashboard.
A rapidly scaling Creator use case had inherited a desktop-first dashboard. Small-viewport usage was growing faster than the product adapted, so onboarding was rebuilt around it.
Generative research with the Creator segment, data analysis and journey mapping; technical validation alongside design so the journey was buildable; hands-on flows, prototyping and usability testing as requirements evolved.
As features kept evolving I stayed on as knowledge guardian, coaching other designers so the approach held up past launch.
A growing Influencer and Creator user base was reaching the platform on their phones, but the dashboard had been designed desktop-first. I reworked it for compact viewports so the numbers a Creator checks most, such as sales, commission and campaign performance, lead each screen. Cards, charts and lists were redesigned to fit a narrow screen rather than simply shrunk to it.
New Creators now start on the same mobile-first footing, with a short step-by-step flow for business set-up, business model and social channels. Social platform authentications were also integrated to suit the persona and use case.
Before any screens were polished, I mapped how onboarding would connect with an Influencer Marketing Platform: from an advertiser inviting an influencer, through registration and verification, to the first offer and a shared link. Flows like this kept design, product and engineering aligned on the whole journey.
Customers buying smart-home technology or choosing a phone and plan often stall at the same point – they aren't sure which option is right. Guided experiences for EDF Energy's Smart-home Store and a machine-learning advisor for telcos.
Choice paralysis is a common occurrence for consumers. A traditional product grid or a page of generic advice doesn't often resolve that. The core challenge across these two commerce platforms was to guide people to a confident choice.
The goal was to increase sales of smart-home technology, helping energy customers use energy more economically and take steps towards a low-carbon future. The barrier to heat pumps and smart thermostats wasn't awareness – it was the intimidation of installation. We learned that generic advice doesn't change habits either. Collaborating closely with the PM, I mapped user needs and identified barriers. I designed modular tips and product recommendations tailored by usage and context, using sales psychology framing. I prototyped and tested behaviour nudges and savings insights directly with users.
For telco companies including O2, EE, Verizon and Vodafone, I worked alongside a cross-functional team to produce an AI advisor that helps customers choose the best phone and plan for their needs. Machine learning was core to the product in improving those recommendations. The work ran from research, workshops, ideation and usability testing through to technical validation, in close partnership with Data Science on A/B tests that challenged the design hypotheses. I introduced agile design while managing a small team, with the assistant supported across native and web.
My career began in visual and brand design, so Design Systems are where that craft can scale across products. A comprehensive UI kit, a one-page design language cheat sheet and an AI copywriting plugin that keeps the product team's tone and language consistent.
The Blueprints UI kit is a comprehensive library: typography, colour and icons, form inputs, buttons and cards, application headers and footers, modals, tables and list views, alerts and status tags and a responsive grid with breakpoints. It underpins both Blueprints Web and Blueprints Agentic. I inherited it as a single-product library. Working with and guiding a team of designers and engineers, we scaled it into the enterprise source of truth, in code and in design. The Product team own the governance, the React and Figma parity and the accessibility standards, plus the contribution process that lets teams extend it without fragmenting it.
A reduced, one-page cheat sheet distils the system to its essentials: colour palette, typography, core components, visual language and iconography, anchored in three words: Bold. Dimensional. Direct. It gives teams a quick reference without opening the full library.
An AI-powered Figma plugin I created for the product team, to keep tone and language consistent across the product. A Generate tab produces multiple copy options from a short description; a Scrutinise tab rewrites existing copy for tone, length and clarity, trained on Rakuten copy style. Any option can be applied straight to a text layer.
Met the Product team’s annual objective to unify content across the product.
Music theory and production background turned into a working melody-generation tool for musicians. Personal project using AI to plan, design and build in iterations. Fully end-to-end process.
MVP hardcoded against real music-theory rules plus a pool of reference motifs. Works fully offline, no server dependencies, fully responsive.
Strictly not AI sound generation. Melodia produces MIDI data only, which people take away and play through their own sound sources. That distinction was a deliberate stance on AI-generated music. It is fully responsive, so it works across all devices.
Stack: React, TypeScript, Vite, Tailwind, Tone.js.
Roadmap: Continuous training, payment gateway, affiliate links to sound sources.
I began my professional career as a graphic designer and art director. I continue to design to this day, but more as a sideline. These are a handful of those sideline projects. Strictly no AI used.
About
UX leadership with a creative core.
UX leader and product designer driving user-centred design, strategy and AI transformation. As AI changes both how we work and the products we build, I collaborate with cross-functional teams to deliver scalable, smart product visions. I ground the work in design thinking and data, always keeping customer experience and business impact in the same conversation.
I mentor, grow and retain design teams through shared vision and instilling empowerment. With product and engineering I put the habits in place that make quality repeatable, staying close enough to research, prototype and measure so we can jointly understand the material we're shaping.
How I work
Work out the job before drawing the screen
Before any design starts I want to know which problem is worth solving and what good results could look like. That means talking to the users, the people doing the work, looking at the data and agreeing the measures with the team who will build it.
Cross-functional team collaboration from the start
Product, engineering, data and research join while the design is still rough. Each of them sees options I wouldn't. We prototype and test early, ship in small steps, then watch how it's used and keep improving it. Launch is where the learning starts.
Give people real ownership
I grow designers by handing them proper work and being straight about how it's going, then helping them towards whatever they want next. I stay with the work through delivery and after launch, coaching the people who pick it up, so the thinking outlasts my part in it.
We experiment with the tools together
Using AI as a tool is now routine. From making sense of research, prototyping and reporting. That includes the shift towards design systems that AI helps to build and maintain. I encourage my team to experiment and share their findings too. Staying hands-on matters when I'm leading a team through the change.
Experience
Leading design for a connected product suite and its design system, with a rapid shift toward agentic AI. Driving AI-powered automation and personalisation, growing and mentoring the design team, and moving legacy ways of working to AI-first. Led an initiative that increased network gross sales by $200M.
Design strategy and technology innovation to help speed the shift to a low-carbon society, with every piece of work tied to growth-driving KPIs.
Led design for an AI-led advisor that boosts digital sales and customer experience, blending sales psychology, machine learning and UX. Led the design team, embedded agile design and worked closely with Data Science on A/B testing and validation.
Led user-centred design across AV-connected applications, hardware and live-stream services, driving agile optimisations and technology solutions. A fast-paced agile working process with frequent usability testing. I also rebranded the product suite.
Created connected experiences and digital transformation through user-centred design at a multi-channel digital technology consultancy. Industries included finance, automotive and energy.
Worked with brands, artists, NGOs and organisations on packaging, campaigns, editorial, merchandise and video, from concept to production.
Editorial design, advertising campaigns, illustration and branding.
Led creative design, branding and campaign briefs from concept to delivery.
Making music has always been central to my sense of creativity.
I continue to write and produce.
Let's talk.
Hiring or want to talk about a project? Send me a note and I'll get back to you.
I'll reply to the email address you supplied.