Pavel Li

About

About

Get in touch

Telegram

LinkedIn

Multi-D

Multi-D

Rosatom Engineering Division

Website

Award

I designed the planning core of the system used to build nuclear power plants: a work tree holding over 100,000 records, a Gantt view, an Excel-like calendar grid and a detail card. Then I turned its components into the design system for all five Multi-D products.

I designed the planning core of the system used to build nuclear power plants: a work tree holding over 100,000 records, a Gantt view, an Excel-like calendar grid and a detail card. Then I turned its components into the design system for all five Multi-D products.

Years

2021–2025

Role

Senior Product Designer

Scope

Research, IA, UX/UI, Design system, Mentoring

CONTEXT

A nuclear power plant takes close to two decades to build. Every physical act on that site, pouring a slab, accepting a batch of rebar, closing out a month of contractor work, is one record in a plan that runs to six figures.

Multi-D is the system that holds it. Five products, five product teams of thirty to fifty people, and more than a thousand users across every active construction site Rosatom runs worldwide.

CHALLENGE

I could not design any of it until I understood the work.

Not the software, the work. What a scheduling methodologist actually does on a Tuesday. Why a site engineer trusts one number and argues with another. What happens between a contractor reporting progress and that progress becoming a fact someone will pay for.

None of this is written down anywhere you can read it.

HOW I LEARNED IT

I went and asked, repeatedly, for four years.

I built a standing line to the people who run the work: weekly calls with the ten most active users, plus their managers and the methodologists who define the rules. Not a research sprint with a deadline, a habit. Every week, the same people, long enough that they stopped performing for me and started complaining honestly.

Then I flew to Egypt, to a live nuclear construction site, to meet them where the work happens. An hour on site answers questions a year of calls will not, because on a call people describe the process they are supposed to follow, and on site you watch the one they actually use.

I went and asked, repeatedly, for four years.

I built a standing line to the people who run the work: weekly calls with the ten most active users, plus their managers and the methodologists who define the rules. Not a research sprint with a deadline, a habit. Every week, the same people, long enough that they stopped performing for me and started complaining honestly.

Then I flew to Egypt, to a live nuclear construction site, to meet them where the work happens. An hour on site answers questions a year of calls will not, because on a call people describe the process they are supposed to follow, and on site you watch the one they actually use.

WHAT I FOUND

Everyone was already working in Excel.

Not as a workaround they were embarrassed about. Excel was the real system: the reporting, the daily routine, the arguments, all of it lived in spreadsheets that people had been fast in for years.

The obvious move is to replace it with something better designed. The correct move was to keep it. I kept the pattern and moved it inside the product, because asking a thousand people to unlearn the muscle memory of a decade is not a migration plan, it is a way to guarantee they keep using Excel and start hating you.

Everyone was already working in Excel.

Not as a workaround they were embarrassed about. Excel was the real system: the reporting, the daily routine, the arguments, all of it lived in spreadsheets that people had been fast in for years.

The obvious move is to replace it with something better designed. The correct move was to keep it. I kept the pattern and moved it inside the product, because asking a thousand people to unlearn the muscle memory of a decade is not a migration plan, it is a way to guarantee they keep using Excel and start hating you.

The work tree

Every record is a real thing happening on a site: a pour, a delivery, an acceptance. There can be more than a hundred thousand of them at once, nested five levels deep. The tree is the spine of the product, and everything else is a way of looking at it.

The Gantt view

The same data as time. Dependencies, durations, and the question every planner is actually asking: if this slips, what else does.

The calendar grid

Excel, honoured. Keyboard navigation, range selection, filling down a column. Every departure from spreadsheet behaviour had to earn itself, because each one costs a planner seconds, thousands of times a day.

The detail card

One work item, everything that matters about it in one place. The dashboard for a single record, for the moment someone stops scanning and starts deciding.

THE SMALL THINGS

Then two years of the work nobody puts in a portfolio: the context menu, the header, bulk actions, the reference books. Improvements measured in seconds saved per interaction, which at a thousand users adds up to the only kind of number that means anything in enterprise.

By then the product was better and, more importantly, the channel was open. Weekly calls with real users, my own seat in team planning and retro. I stopped being the person who receives requirements.

Then two years of the work nobody puts in a portfolio: the context menu, the header, bulk actions, the reference books. Improvements measured in seconds saved per interaction, which at a thousand users adds up to the only kind of number that means anything in enterprise.

By then the product was better and, more importantly, the channel was open. Weekly calls with real users, my own seat in team planning and retro. I stopped being the person who receives requirements.

THE SECOND LIFE OF THIS WORK

The planning core needed components that existed nowhere else: dense editable tables, timeline elements, tree nodes. Rather than draw them for one product, I built a universal component base on top of Ant Design for all five Multi-D products.

I rolled it out gradually rather than by decree. Teams took it because it was faster than building their own, which is the only kind of adoption that survives the person who started it.

Alongside it I mentored the junior designers and ran a design workshop every Friday.

OUTCOMES

1000+

Users worldwide

100000

Records in a plan

5

Products on one design system

Four years, one product team of my own and four more that took my components. I left the system in a state where the people who use it can argue with the designers who build it, which was not true when I arrived.

Four years, one product team of my own and four more that took my components. I left the system in a state where the people who use it can argue with the designers who build it, which was not true when I arrived.

Back

THANKS FOR READING THIS FAR

I am looking for my next team, Senior or Lead, remote or Saint Petersburg.

I am looking for my next team, Senior or Lead, remote or Saint Petersburg.

Get in touch

Telegram

LinkedIn