Roman Leonenko

I started Interface Lab to write about what happens after an AI idea starts to work. A demo can show the promise; the product work is making that promise understandable, inspectable, and useful in real workflows. The posts here focus on the frontend side of that work: how model output becomes UI people can check and use, where human judgment sits in the loop, and what changes when data is shaped by an LLM.

A good demo is a meaningful milestone. It proves a direction, creates momentum, and helps a team learn quickly. The next step is different: a product has to support everyday users, odd inputs, changing teams, and the expectation that important behavior stays clear over time. Much of that step is interface work.

Background

I’m Roman Leonenko, a frontend developer and consultant based in Munich. For thirteen years my work has been enterprise digital platforms in Europe — automotive, e-commerce, and consulting — including a Mercedes-Benz-related platform built from the ground up and e-commerce projects connected to Mars and Royal Canin.

That work has meant complex business logic, long-running delivery, technical ownership, microfrontend architecture, and, more recently, practical use of AI-assisted tooling in everyday frontend teams. Stack: Vue, React, TypeScript, microfrontends, and component architecture.

What shows up here

I write about how model output becomes structured UI, how component systems handle generated content, how review steps make AI actions inspectable, and how protocols like A2UI and AG-UI are changing the frontend side. I try to keep it practical: patterns that still make sense once a team has to maintain them.

A working model is a milestone; the interface is how it becomes useful. Chat is often a good entry point, and it becomes stronger when combined with structured UI, real components, and clear boundaries between what the model proposes and what the product commits to.