From LMS to streaming service
Rebuilt a Russian cultural-education platform — 157,000 users, 100+ courses — from an ageing LMS into a streaming product, then shipped iOS and Android off the same system in ten days with no additional design work. Purchase conversion +32% on day one, D7 retention 23% → 30%, authorisation 65% → 91%
Role
Product Designer
Industry
EdTech / Streaming
Year
2025
Platform
iOS, Android & Web
Context
Strong content, inside a product people avoided
The platform had what most education companies want: 157,000 users and a library of more than a hundred courses that people genuinely rated. What it did not have was a product that let them reach it. The content lived inside an ageing LMS whose interface had stopped scaling, and the numbers said so without ambiguity — a sticky factor of 7%, an average of two visits a month, and a customer lifetime that ended after one or two payments.
The product owner ran depth interviews because the decline was starting to look terminal. What came back had nothing to do with content quality. People described navigation they could not hold in their heads, missing basics like watch history and recommendations, and a feeling of having bought one thing and received another — not because the lectures were weak, but because getting to them was work. The more technical among them had counted the clicks between opening the app and the thing they had paid for. Watching on a phone was not a scenario anyone attempted.
I worked as the product designer on the engagement, alongside the client’s product owner, their in-house designer, and the agency’s team and developers. My scope ran across the product sprint, the visual language, the mobile adaptation and the design system underneath all three.
Challenge
Constraints
— Two months for design. Six for the whole project including development.
— Web development ran in parallel from day one, so every idea had to survive in a browser as well as a hand.
— The client had been burned by a previous studio and started openly sceptical of agencies. Trust had to be earned with work, in the first week, not with a pitch.
— No unified visual language existed at brand level. There was nothing to extend, so it had to be built from zero inside the same two months.
— No separate budget or time for iOS and Android design. Whatever mobile got, it had to inherit.
— The system had to survive handover. Anything the in-house team could not maintain alone would be a liability, not a deliverable.
Discovery
The users had already counted the clicks
The interviews had done the diagnosis. The question left was what kind of product the answer implied. Every complaint pointed the same way: people were not asking for more courses or better lectures. They were asking for the behaviour they already had everywhere else they watch things — open, continue, next.
So the first week went on taking those products apart. Netflix, HBO Max, Kinopoisk, MasterClass, YouTube, short-form social video — not for how they look, but for the mechanics of moving someone from open to playing, and from one thing finishing to the next thing starting. That teardown, delivered inside a week, was also how a sceptical client found out whether we were worth working with.
157,000
users, across a library of more than 100 courses
7%
sticky factor — the share of monthly users showing up on any given day
2x
average visits per month, per user
1-2
payments before a customer’s lifetime ended
Streaming teardown: Netflix, MasterClass, Kinopoisk, HBO Max
New Design System
Core concept
Eleven ideas, none of them from education
The sprint produced eleven proposals. The ones that survived were the ones that removed a decision rather than adding a feature. Continue Watching moved into the header. Tabs replaced infinite scroll. Recommendations shipped without AI — a deliberately simple algorithm, legible enough to tune and cheap enough to fit the timeline. Short video became the way into a long course. Progress moved into the sidebar, and a lecturers section and light gamification came in behind them.
01 - Continue Watching, In the Header
The most-wanted action on the platform stops being something you navigateto.
02 - Tabs, Not Infinite Scroll
A library you can hold in your head beats one you can scroll forever.
03 - Recommendations Without AI
A deliberately simple algorithm — inspectable, tunable, and shippable inside two months.
04 - Short Video As The Way In
Thirty seconds to decide whether a twenty-eight-hour course is for you.
Custom icon set drawn from scratch & paintings used as avatars and in-interface material
Light as knowledge, made literal
There was no brand visual language to extend, so we ran a workshop with the founders and asked what the platform actually is to them. Three metaphors came out of it and became the brief: light as knowledge, the magic of discovery, gallery space.
What that turned into is specific. A custom icon set drawn from zero so the same idea reads the same in every corner of the product — the home icon is a temple of the arts. Paintings from the library’s own subject matter moved into the interface as banners and avatars, which gave the product depth and tied the surface to what it teaches. And the like stopped being a click: it dissolves into stars, because what a good lecture does is click in your head.
The navigation rebuilt: filter, resume, continue
Cards that take their color from the video
The card backgrounds are not fills. Each one is assembled from the video it belongs to: the key frame the user sees on top; a second copy of that frame underneath at a different crop; that layer blurred heavily into a soft colour bed; then a darkening pass so the text stays readable over any of it.
The result is a surface that always matches its content and never fights the image sitting on it. The rule is identical on mobile — which is part of why mobile cost nothing to design later.
The multi-layer card
Validation
Re-invented player
On desktop the player followed the YouTube model. On a phone there is not enough room for that, so it was rebuilt instead of scaled down. Timecodes moved directly under the video. Navigation collapsed into two fixed buttons — "All lectures" and "Next" — so you change lesson without leaving playback. Like and dislike are pinned, so reacting never costs you your place. Previews of neighbouring courses appear inside the player itself: finish one, see what follows, search for nothing.
Three more decisions came from how the thing is actually used. Autoplay, because lectures get listened to with hands full — cooking, cleaning, driving. Completed sections collapse into an accordion, so a thirty-video course is not a scroll. And a random block for the case nobody designs for: wanting something new without knowing what. Press it, get a course, take it or roll again.
The mobile player with timecodes and adjacent-course previews
The things only the app could do
PiP, so a video shrinks into the corner and browsing continues around it. That is convenience, and it is also straight forward time-on-platform, which is the metric the whole project started from.
Offline download, which is the strongest retention feature in the set. Learning that does not depend on a signal is learning people come back to — on a plane, on a commute, anywhere the connection gives up.
And cross-device continuity: start on the web, resume on the phone, with progress, downloads and settingsintact. The app never announces itself as a second product; you are just still where you were.
The system outlived the engagement
For the build I assembled a UI Kit that became the single source of truth: tokens for colour, type and spacing, a component library, and the rules for using them. The architecture was designed from the start to scale across web, mobile and marketing rather than being retrofitted to them.
That is what made mobile cheap. iOS and Android came together in seven to ten days including development,and the design cost zero additional hours, because every adaptive rule had already been decided.
At handover the whole base went to the in-house team. A year later they had used it to rebuild the personal account area, refresh the front end, and bring external communications into one style — including an interests-based onboarding flow built entirely without us. A system that keeps growing after its authors have left is the only real test of one.
7–10 days
to ship iOS and Android — design and development together
0 hours
of additional design for mobile — the adaptive rules were already decided
2 months
for the full product design, inside a six-month project
1 year on
the in-house team was still shipping new features on the system
PiP, offline and the app in situ
Impact
What the numbers did after release
Delivered on schedule in March 2025. Purchase conversion rose 32% on the first day and held at +25%across the week. Authorisation climbed from 65% to 91% inside a week. D3 retention moved from 33% to 37%, D7 from 23% to 30%, and engagement added a further 5–8%.
D7 is the number that matters here. Day three tells you the release was not broken; day seven tells you a habit is forming. Going from just under a quarter of users to just under a third at a week is the difference between a library people own and a product people use.
The sticky factor the project started from was never a content problem. Once continuing was easier than deciding, people continued.
Other projects
2x conversion on the mobile homepage
Redesigned a top-10 Russian bank's mobile homepage from discovery to design language. Validated with eye-tracking and a staged A/B rollout: product-showcase conversion 4.2% → 8.14%, with no metric regression.
From one store to an omni-service
Turned one of Russia’s grocery chains’ app from a hypermarket delivery service into an omni-app covering five store formats. A/B tested to 30% of traffic: format switching 0 → 1.62%, loyalty engagement +25–30%, offline ARPU +1.99% — against a 1.36% dip in order conversion.
TikTok Creators: Safety Tools
I turned enforcement-quality data into structured product recommendations, built the policy-to-product layer underneath safety surfaces, and designed the internal tooling and measurement moderation operations ran on. The direction that evidence argued for — prevention at the point of creation rather than review after it — is the one TikTok’s 2025 creator-safety tools follow