A robotics marketplace, designed alone in three months

Context: UXdevice, a marketplace for robots, built from the client’s idea Role: the only designer, freelance for a private client When: 2026, about three months of design Live site: uxdevice.ru

Before the project the client had only an idea: a marketplace for robots. He wanted the design and the site fast, to launch a startup.

According to the client, the buyers are mostly companies that use robots in production and in marketing, and some of them resell. There is no payment on the site. A buyer finds a model and sends a request with contact details, a manager calls back, and my part ends there.

I did the UX, the UI, the design system and all graphics, including the logo. The client and I had a call once or twice a week, and the last word was his.

First I proposed a bright concept with custom graphics and animations, but the client said no. He saw the main buyers as people over 40 and wanted a minimal site with simple logic. A plain site was also faster to build.

The interface is plainer than I wanted. In return, I put the freed-up time into the UX and left room to improve the look later.

The client wanted to split the catalog into pages. I argued for one list with a button that loads the next models.

A buyer who opens the catalog often does not know yet which model they need, and one growing list is faster to browse. With pages, the buyer cannot recall which page had the robot they liked earlier. The client agreed, and the button is on the live site.

For the catalog I drew ten scenarios: filters before and after they are applied, sorting, a category view, and search with and without results.

I added a third level of tokens to control component states in one place, without opening the pages.

The design system has 831 variables in three collections: 188 primitive tokens with the raw colors, 229 semantic tokens for roles such as background, text and stroke, and 414 component tokens, each for one component, such as a button or a tab. Semantic and component tokens point only to other tokens, never to a raw value.

On top of the tokens are 40 component sets, from buttons and inputs to the product card, the header and the price filter.

A studio built the site. I checked their pages against my layouts and wrote each issue as a card with three fields: what is wrong now, how it looks in the layout, and why.

There were three rounds, with 25, 8 and 3 cards, 36 in total. A card shows the site next to the layout and has a priority. For example, the top bar of the header disappeared on scroll and did not come back. My card said that scrolling up must bring it back, on every page.

One card is still in the “Not fixed” list.

I used Claude and Grok to review competitors and references, and for most of the interface copy. I checked the results and looked at the sources myself. I drew every layout myself.

On 7 October 2026 the sitemap of the live site listed 39 products in 4 catalog sections (my count). The catalog with filters, the product pages, comparison, favorites and rental are all on the live site.

I had about a week and a half for all the mobile layouts, and they have UX mistakes. The main problems are in navigation, both in the header and in the footer. Next time I would first study references, build a visual concept of the mobile version, and only then carry the patterns to the other pages.

The second gap is the price filter. I designed a slider with fixed steps, and the studio built one without steps. I reported it, and the client kept it for the MVP.

For a scope like this I would now plan one and a half times as much time and ask for an analyst and a second designer in support.

Next case: Six versions of one screen: a walk-to-earn game inside Telegram

All work · CV · hello@eugenesolomatin.com · Telegram