[ § 01 ]
About
I'm a product designer working on marketplaces and complex platforms — the kind of products where the hard part isn't the screen; it's deciding what should be on it. Most of my work sits around discovery, trust, and conversion: how people find things, how they decide whether to trust them, and what stops them. I work research-led and hypothesis-driven, which mostly means I'd rather find out than argue.
[ § 02 ]
How I work
- I start from evidence, not opinion — but I'd rather make a decision than collect data forever.
- I work closely with PMs, engineers, and data. Design that ignores constraints isn't design; it's decoration.
- I'm comfortable with ambiguity and I'll say when something isn't yet answerable.
- I test what I believe, and I'll tell you when I was wrong.
[ § 03 ]
Principles
- Evidence over opinion — but a decision over endless evidence.
- Clarity is a feature.
- Constraints are the brief, not the obstacle.
[ § 04 ]
How I work with AI
I use AI where breadth and speed help, and I check it against my own judgement before anything ships.
On discovery, I gather the raw insights myself and use AI to synthesise them — I run a structured skill (Atomic PMF) to pull findings into themes, then go back through it against my own read of the sessions. The synthesis is faster; the interpretation stays mine.
I use it to build, too. I started a redesign of my own Shopify store through Claude Code with the Shopify connector (it's paused for now), and I use ChatGPT to generate visual mockups for products before committing to a print. And I use it to stay current — accessibility standards and practice move faster than I can read.
Where I override it: taste, and any judgement about my own work. While I was building this site, it recommended an accent colour I disliked on sight, and it sequenced the build so visual craft came last — efficient on paper, demoralising in practice. Both times it was reasoning correctly from what it knew. It just didn't know the thing that mattered.
At work, it's part of how I design. On Quality Label I built the iterations with Claude Code and the Figma MCP: more than 30 versions in Figma, and the ones worth testing as working HTML prototypes, so teachers tested the real thing.
It needs checking in research too. On the search discovery, Claude was my synthesis partner, and it split the Greek and German teachers into separate populations. I pushed back, because intent doesn't depend on the market, and intent was pooled across all seven. A QA pass then caught four overclaims before anything was shared.
Designing AI-powered products
At eduki, an AI assessor scores teaching materials against a quality framework. I'm designing what buyers see of that judgement: a module that says what each material actually does well, tested in rounds. There was no prompt yet, so I wrote the rules a generator would have to follow: only criteria rated high get a line; no scores or verdicts; every line names something concrete in the file. A fourth rule, "one fact, one line", came out of a designer's review in round 1, and it's now in the PM's requirements for the prompt.
Designing for a model also means knowing where it's inconsistent: the assessor can rate the same criterion differently from one run to the next, so a line can appear and disappear. That went to the PM as a risk to solve before anything ships.
Sometimes the job is to test the bet first. Before anything was built for eduki's Q3 bet on AI-assisted search, I ran a discovery to find out whether teachers actually want to search in sentences. None of the interview participants did, and the bet didn't go ahead. What they needed was help with the short queries they already type, and that became next-word search suggestions: I designed the feature, the suggestions come from the search backend, and it shipped.
[ § 05 ]
Where this started
DiverBubbles came before the design career, not after it. I started it because I wanted to build something of my own — a side business. Doing it meant doing everything: the brand, the products, the store, the site. That's where I found the part I actually loved — the flows, the interfaces, the decisions about how something should work — and it's why I moved into product design. The store is still running, and I still design for it.
[ § 06 ]