Viktor Bezdek

I lead consumer-facing engineering at Groupon — web, mobile, the whole stack a customer touches. Part of the job is figuring out how a legacy e-commerce platform becomes AI-native without breaking the things that already work. That means picking which workflows to automate first, building the platform layer that makes agents safe in production, and rewiring how engineering teams plan, hire, and ship.

Concrete results from the first year: cut time-to-market by ~30% in three months by fixing org design and killing dead processes. Rolled out AI tools across planning, coding, QA, and analytics — gave my teams back 100+ engineering hours a week. Built an internal prompt optimizer over our tools, systems, and tribal knowledge. Pushed AI from "experiments" into shipping features and measurable cost savings.

It's hard, partial, often two steps forward and one back. The articles here are what I write down when I'm trying to think clearly about a specific piece of it.

AI Native Builders is my working notebook. Each article starts as a question I'm wrestling with at work, a pattern I've seen elsewhere that seems to actually move the needle, or a piece of vendor BS I need to argue with. I write them down so I think more clearly. I publish them so other people working on the same problems can disagree, steal what's useful, and tell me what I got wrong.

The frame I keep coming back to: how do you transform a legacy company into an AI-native one without burning trust, breaking customer experiences, or producing more theater than ROI? I don't think anyone has cleanly solved this. I'm trying to.

No-nonsense. No buzzwords. Just what's actually working, what isn't, and why.