All work

Mabrook

A Dubai super app for video, live, music, games, shopping, hotels and an AI assistant. I led design as the product grew from a scattered MVP into one system.

Role
Head of Design, via Valor Software
Team
3 senior product designers including me, one of them hired by me; graphic and motion designers; product team grew from 15 to 32
My part
Led the design team, the navigation rework and the design system, and designed features myself: AI chat, dating, music battles and challenges, the news ticker and Explorer
Platforms
iOS and Android, built in Flutter
4×monthly active users
9% → 14%day-7 retention
11% → 19%of monthly users open the app daily
10 → 4 daysfeature release with AI-assisted design

Before I joined, Mabrook had spent a year in slow development. The previous development team had built much of it with hardcoded, static screens that only looked like working features, and the build often drifted from the design. There was no design system: random sizes, fractional font sizes, mockups drawn like banners. Oleksandra, the only designer on the product, had started a design system. Every new vertical still arrived with its own patterns.

With food, hotels, shopping, music, games and live streams in one app, the real problem was navigation and a design system that could keep scaling as features kept coming.

  • Team. Joined Oleksandra, saw that one more designer was needed, ran two interviews and hired a senior product designer. The three of us aligned on one direction, with marketing graphic and motion designers alongside.
  • Process. Set up critique, reviews and handoff to Flutter engineers, and ran feature demos for engineering, the founder and management.
  • Design system. Took over the system Oleksandra had started and made it the shared context for the team, and later for AI-assisted design.
  • Navigation. The horizontal top tabs could hold the verticals, but menu items kept growing. We merged some, moved others and reworked the bottom bar, and took each change through the CPO, engineering and the founder with flow diagrams and the reasons behind them.
  • Product decisions. Where requirements were thin, I wrote the missing details into engineering tickets myself. For the news aggregator I defined the categories, explained why, and got them signed off.
  • Leadership. Joined the weekly sessions with the founder and the heads of engineering while the product team grew from 15 to 32 people.

Hypothesis. We believed that if every vertical lived in one feed with one search and one checkout, people would try more sections instead of getting lost between separate patterns. We checked it in Amplitude over the six months after the January release: monthly active users grew 4× and daily stickiness rose from 11% to 19%.

  1. Reset the foundationRebuilt the visual style and many core features almost from scratch instead of patching the old UI.
  2. One feed for every verticalA single full-screen feed with 14 top tabs, one global search and one contextual action: Order now, Book now, Buy or Subscribe.
  3. One checkoutGoods, food and stays share one flow, with a 15-minute booking hold and three ways to pay.
  4. One creation pathPost, reel, podcast, story and live start from the same + button and share one short path to publish.
  5. Retention through playLeaderboards, referral codes, in-app games and weekly music battles.
  6. A system that scales25 component sets, 470 variants and states, 160 icons, with AI-assisted design cutting feature releases from 10 to 4 days.

The AI assistant sits in the centre of the tab bar: voice or text, in the user’s language, with answers read aloud. A safety layer is on the roadmap: 3 age modes, a 3-hour daily feed limit and 18+ for DMs and live.

Inside the team, AI-assisted design took a feature from brief to release in 4 days instead of 10.

  • Monthly active users grew 4×.
  • Day-7 retention rose from 9% to 14%, day-30 from 3% to 5%.
  • Daily stickiness (DAU/MAU) rose from 11% to 19%: almost one in five monthly users now opens the app every day.
  • 16+ verticals run on one navigation pattern and one checkout.

Engagement was a signal, not the goal. What changed for people: food, stays and shopping sit in one feed with one search and one checkout, instead of a new pattern for every vertical.

How we measured: in Amplitude over the six months after the January release. Results reflect the work of the whole product team.

  • Video feed
  • Live
  • Music battles
  • Games
  • Shop
  • Food
  • Hotels
  • Explorers
  • AI assistant

Get into the room where ideas start. New features reached design through the CPO as long, AI-written requirements: a lot of text, little of it about the user. I pushed for better documentation, but the real fix was earlier: being part of the conversations with the founder while an idea was still forming. The closer you are to where an idea is born, the more you can shape it.

One example: we built a dating feature quickly from the first requirements. At the demo the founder reframed it after three screens: questions first, photos only after a mutual match, made for long-term relationships rather than quick dates. Had we shaped that together at the start, we would have saved a full cycle.

Next caseASI Central · ESP+Print-on-Demand, the Cosmos design system and AI-built prototypes in code: brief to prototype 11 → 3.5 days.

Need someone to lead
your
design team?

I’m looking for a Head of Design role, remote from Spain. English, Ukrainian and Russian, intermediate Spanish.