Lidiia Mialkina

In the app catalog, a buyer struggled to find green coffee for their needs on their own and often had to ask a manager to place the order. Goal: raise self-serve orders from 10–20% to 30–40%

Framed JTBD and a hypothesis for a new catalog structure, then in ~15 active hours built a clickable prototype in Cursor — 8 iterations: shelves on the left, filter chips, sticky tools, denser lot cards with cart. Goal: validate UX myself without waiting on external engineering

Status: prototype ready for testing with 5–7 buyers — self-serve order lift not measured yet. Success criterion: 5 of 7 complete a typical order without a manager. Engineering will get an aligned scenario and spec — not a raw hypothesis

The team had not approved mockups for a new logistics flow — map, ETA, and cart hierarchy was contested. Handing a raw spec to engineering meant a high redo risk

In ~2 working days I built a clickable prototype in Cursor (HTML/CSS/JS) so the team could click through the path before development. Consciously in scope: map, ETA as forecast, tracking from cart. Cut: Yandex map underlay, third-party ports, fullscreen by default. A prototype for validation, not a production release

Stakeholders walked from lot card to tracking on a live prototype. Aligned the hierarchy before engineering started — they got a tested spec, not a raw hypothesis. Cut the UX alignment cycle from ~2 weeks (mockups and meetings) to 2 days

On the curated-prompts screen a new user does not see that the query list drives auto-competitor selection and the audit baseline. First value arrives skewed — numbers are computed around the wrong competitive set

In 2 days I built a CJM to first value, a blocker map prioritized with RICE, chose one improvement, and wrote a backlog spec plus a clickable as-is / to-be prototype on Pixel Plus UI Kit tokens

A ready report for the backlog: focus — explain the prompts → auto-competitors / analytics chain before Create (RICE 12.2). Prototype with a Now / After toggle