LOADING 0%
[ Contact ]

PAISAK4U — AI product interface consulting by Katarzyna Pilarz

Designing and engineering digital products

[ UX/UI · React · AI product design ]

[ Helping product teams turn complex requirements into clear, usable interfaces ]
[ Designing for clarity, predictability and trust — especially when the product is complex ]
[ About ][ Independent practice — est. 2024 ]

Product UX and frontend engineering.

I'm KATARZYNA PILARZ, and PAISAK4U is my independent consulting practice, specialising in AI-native product interfaces and UX-driven frontend engineering.

I work across the full delivery cycle — research, system design, UX testing, and frontend — aligning product managers, designers, data scientists, and engineers around what to build.

My focus: turning complex requirements, technical constraints, and emerging technologies into interfaces that feel clear, useful, and considered.

7+ years building digital products, including 2 years shipping GenAI in production — pairing product thinking, design, and engineering with delivery rigour.

Currently researching how Lean Six Sigma can reduce late-stage rework in GenAI MVP delivery — validating the right thing before too much is built.

A man in an armchair with a Dispatch job page open on his laptop: client details, the job’s six steps, its schedule, services and invoice
[ DISPATCH — A FIELD-SERVICE PLATFORM DESIGNED AROUND THE LIFE OF A SINGLE JOB ]
[ How I think ][ Six principles ]

The solid rules behind each interface I design, build and test.

  1. [ 01 / 06 ]

    Aligning user needs with business goals

    Every interface decision should serve both the people using the product and what the business needs it to achieve. I start by agreeing those goals and how success will be measured — so the work is judged on outcomes, not output.

    [ “Validating the right thing before too much is built.” — Research ]

  2. [ 02 / 06 ]

    Systems before screens

    Before designing an interface I model what sits underneath it: actors and responsibilities, states and transitions, data relationships, business rules and failure conditions. The interface is the visible layer of a system.

    [ “A scan is context. A basket line is an identity.” — QRCodeFlip ]

  3. [ 03 / 06 ]

    Predictability over the happy path

    I design what happens when the system changes — stale data, an item that disappears mid-order, an interrupted stream, a result that needs verifying. People should always know what changed, what is happening and what they can do next.

    [ Menu changed → basket reconciled → guest told → checkout held — QRCodeFlip ]

  4. [ 04 / 06 ]

    Design that survives implementation

    I don't stop at the design file. I think about behaviour, edge cases, accessibility and implementation while designing, then stay close to the code until the experience actually works.

    [ Figma, tokens, components and accessibility share the same rules — QRCodeFlip ]

  5. [ 05 / 06 ]

    Evidence over assumptions

    Decisions are checked against prototypes, real users, browser testing, accessibility audits and measured behaviour — and a fix lands where the rule lives, not just where the symptom shows.

    [ Contrast computed from the tokens actually shipping, not assumed — QRCodeFlip ]

  6. [ 06 / 06 ]

    AI with human accountability

    AI accelerates exploration, implementation and validation. Product decisions, architecture and acceptance stay with an accountable person, and model output is held to the same rules as everything else.

    [ “The model interprets. The product owns the facts.” — uxproof · “The agent finds. I decide. The system validates.” — QRCodeFlip ]

Selected work

[ Selected work ]01 / 02
01Product UX / Frontend engineering / Design systems / Accessibility · 2026 · Completed for v1

QRCodeFlip — a multi-context ordering system

A laptop open on a mustard armchair, showing the QRCodeFlip admin dashboard: today's net sales, orders, menu views, the busiest QR point, recent orders and the top-selling drinks
[ THE ADMIN DASHBOARD — TODAY'S SALES, ORDERS AND TOP SELLERS AT A GLANCE ]
[ Problem ]Many hotels, each with its own menus and locations, on one data-heavy backend. Guests needed QR ordering with no account or app, staff one place to run it — while menus change mid-order.
[ Approach ]I modelled the system before the screens — what a scan means, when two basket lines are the same, what happens when the menu changes — then built both sides in React on one design system.
[ Read case ]
02UX engineering / AI product / Design systems · 2026 · Prototype-Built

Uxproof — engineering reliable AI experiences

A generated uxproof deck on a meeting-room screen: the violet findings-summary slide
[ THE DELIVERABLE, IN THE ROOM IT'S BUILT FOR — THE FINDINGS POSTER, THE DECK'S ONE VIOLET STATEMENT SLIDE ]
[ Problem ]AI speeds up research reporting, but a client deck has to be right every time, not most of the time — and a local model, needed for confidential data, will sometimes be slow, wrong or down.
[ Approach ]I drew the boundary first: the model may interpret, but code owns the numbers, the slide structure and the styling. Typed contracts check every output, and every AI step has a fallback.
[ Read case ]
[ Recommendations ][ From LinkedIn ]

“Turning complex requirements into simple, intuitive user experiences.”

I had the pleasure of working closely with Kasia as part of our GenAI team at DSB, where she played an important role as our frontend developer.

Kasia has a strong eye for UI/UX and a great ability to turn ideas and sometimes quite complex requirements into simple, intuitive user experiences. We worked closely together on several parts of our GenAI platform, where she was always constructive, reliable and willing to challenge ideas when she saw a better way of doing things.

What I especially appreciated about Kasia was her collaborative mindset and the care she put into the details. She takes ownership of her work, is easy to work with, and genuinely wants to deliver something that users will enjoy using.

I would happily work with Kasia again and can strongly recommend her to any team looking for a skilled frontend developer and a great colleague. —

[ Serkan Altay · AI Architect & Lead, GenAI platforms — DSB · Aug 2026 ]
[ Research ][ Ongoing]

How Lean Six Sigma can reduce late-stage rework in GenAI MVP delivery.

GenAI MVPs can be rebuilt multiple times — not necessarily because the technology failed, but because the wrong problem, workflow, or interaction was validated too late. Requirements shift after the demo, edge cases surface in pilot, and interface decisions are revisited in the last sprint.

This research applies Lean Six Sigma — DMAIC, value-stream mapping, defect-at-stage measurement and root-cause analysis — to GenAI delivery cycles to understand where rework originates and which early validation gates can remove it at the lowest cost.

Independent research and thesis towards Lean Six Sigma Green Belt certification, building on postgraduate AI Project Management studies completed in 2026. Findings will be published here as the work completes.

[ Services — Full product UX ]

[ Discovery to pilot · in partnership with your team ]
  • A discovery workshop — mapping problems, opportunities and what done looks like on a whiteboard

    [ 01 / 04 ]

    Discovery

    Short, structured workshops with the people who own the problem — product, design, engineering and the business around one table. Together we clarify users, constraints and success criteria, and leave with the few scenarios worth building and a shared definition of “done”.

    [ User journeys · Information architecture · Human–AI interaction ]

  • The paisak4u design system open in Figma on a laptop: its colour and typography foundations, with the file's pages and layers in the side panel

    [ 02 / 04 ]

    System design

    Designing behaviour, not decoration. Flows, states, components, tokens and accessibility designed together, and prototyped through loading, empty and error states while change is still cheap. For AI products: uncertainty, errors and progress, not just the ideal result.

    [ Figma · Figma Make · Design systems · Design tokens · WCAG / accessibility · Data visualisation · Prompt design · Agentic workflow design ]

  • A development setup — production code across a laptop and an external monitor

    [ 03 / 04 ]

    Frontend

    Design that survives implementation. Production React and TypeScript, built hand in hand with your backend developers and engineers — accessible to WCAG, tested with Jest and Playwright, with streaming AI states treated as part of the experience.

    [ React · TypeScript · Next.js · Astro · React Native · Tailwind · Zustand · Streaming UI · Framer Motion · GSAP · LLM application development · Claude Code · Cursor · Jest · Playwright ]

  • A pilot review — real-user findings and a prioritised change list on a whiteboard

    [ 04 / 04 ]

    Pilot & iterate

    A pilot with real users, not a demo. Behaviour observed with your team, feedback and usability issues turned into a prioritised change list, and the highest-value improvements shipped together in the next release — iteration driven by evidence, not opinion.

    [ Usability testing · Lean Six Sigma · Agile & Scrum · Jira / Confluence ]

[ Questions ][ Before we talk ]

A few practical answers.

Do you work with existing product and engineering teams?

Yes. I have spent 15+ years in international, cross-functional teams, and most of my recent work has been embedded in an existing product and engineering environment — for the past two years in DSB's GenAI team in Denmark, alongside product managers, engineers, data scientists and domain specialists.

I'm used to collaborating across countries, languages and time zones — from an enterprise team to returning international clients — and I can work alongside an existing designer, own a defined product area, or bridge the gap between design and frontend engineering.

Can you help with a product that is already in production?

Yes — and often that is where the most useful work starts. An existing product provides evidence: where users hesitate, which workflows have grown complicated, which components have drifted.

I work incrementally — UX improvements, design-system modernisation, accessibility, interaction redesign or frontend — without the whole product having to stop.

Do you work across both design and frontend?

Yes. I can take a problem from discovery through information architecture, interaction and visual design, then build it in React and TypeScript.

Working across both changes the quality of the design: component behaviour, responsive and loading states and edge cases are considered before development, not discovered during handoff. I can also work on either side alone when that is what the project needs.

How do engagements usually start?

With a short note about what you are building, where it is now — idea, prototype or production — and what you need help with.

Then a 30-minute conversation about the problem, constraints, team and desired outcome. If there is a fit, I define the scope, approach, deliverables and timeline before work begins.

Katarzyna Pilarz
Years building digital products end-to-end
7+
Years inside DSB's GenAI team
2
Years in international, cross-functional teams
15+

Let's discuss
your project

[ BOOK A 30-MIN INTRO CALL — NO COMMITMENT ]