AI-native product design for SaaS and e-commerce teams. Machine-readable design systems, Figma-to-code pipelines, and one human gate before anything ships. One designer, a team's throughput.
Book a callNow it is one designer, a system a machine can read, and agents that assemble the rest. Tokens, components with props and slots, and the rules of what good looks like are written down once, so screens and pages get built from them instead of redrawn every time.
I keep the calls a machine can't make: what the product should feel like, which of forty options is the right one, and when it is done. Everything else runs on the pipeline.
Selected work
Shots (7)
Every shot, with the process behind it, is on Dribbble.
Case studies (2)
Five years of shipping products, from early MVPs to token systems a whole dev team builds from. At VideoPoint the token system cut idea-to-production from weeks to days.
products shipped
brand identities
years in product design
Product UI
Dashboards, configuration, multi-state flows, merchant apps. Clarity inside density, built against a component registry so what you approve is what gets built.
Design systems
Tokens, Figma variables, components with props and slots, documented so developers and agents build from the same source. One system across app, widget and site.
Design pipelines
Claude Code skills that carry the rules of your product, deterministic checks for tokens, contrast and geometry, and one human gate. Most production UI gets assembled; I own the part that isn't.
Marketing sites
Landing pages and sites that are measured, not admired. Built in code, not in a site builder, with the same type and colour system as the product.
UX research
Session replays, interviews, journey maps and jobs-to-be-done. Insight docs written to be acted on, and pipelines that keep the monitoring running after the study ends.
01
One page: who reads it, the sentence they should think in three seconds, the one action, and what is already fixed. Nothing is designed before it exists.
02
Tokens, type roles, components. Machine-readable before the first screen, so every screen after it is derived, not invented.
03
Screens and pages get built from the system by agents running the rules. I make the calls: direction, the right option, done or not.
04
Every batch passes deterministic checks, then one human review: mine, then yours. Then it ships to your codebase.
Free tool
Paste a URL. The X-Ray renders the page, reads every style the browser actually applied and counts what the design system consists of: colours, spacing, type, radii, shadows. Then it proposes the shortest token list that covers all of it.
Why it matters. Every value typed by hand is a decision nobody made: five greys that are one grey, 53 spacing values where ten would do. That is why the same button takes a week to change. The X-Ray puts a number on it and hands you the token scale that fixes it, without changing how the page looks.
Run X-RayRuns in your browser, nothing is stored, no sign-up. You get a ready :root block and a migration map.
15257
hard-coded values on the page, and the tokens that cover them all
Spacing 53→10Colour 46→31Type 23→9Shadow 17→3Radius 13→4
Nick Besiedin, product designer in Tbilisi. Five years across SaaS, AdTech and e-commerce. Now at LifeOS; before that VideoPoint, Advision Digital, Centum-D, and fifteen-plus startups as a freelancer. Since 2025 design here runs the AI-native way: the system is machine-readable, agents do the assembly, and the taste stays human.
Time zone: Tbilisi is GMT+4, a full working day with Europe and mornings with the US East Coast. Portfolio also as a PDF.
or write to nick.besiedin@gmail.com. I read every brief myself and answer within 24 hours.
