Yurii Liubchak

I build products used by millions

I'm Yurii Liubchak, AI lead at KNESS in Ukraine. I take AI from a messy business problem to a system people rely on every day. The millions are mostly in Telegram, in products I've built and run solo for years.

Tell me about a problem → Telegram @LyDev
10+years in production
150+projects shipped
1K+open-source stars

What I'm doing

Main workAt KNESS Group, I find workflows worth automating, build AI agents around live company data and take them all the way to stable production. The job is equal parts research, engineering and getting a system adopted by the people who will actually use it.

Independent workI run fStik, QuotLy and other Telegram products on my own infrastructure, and I build Capka, an open-source AI workspace. I occasionally help companies with difficult AI and automation problems where a polished demo is not enough.

Selected work

More products, experiments and notes · @LyBlog →

About

Started at twelve, editing uCoz sites with too much time on my hands. Then came PHP, then VK bots, then the Telegram ecosystem I still live inside today.

Most of those millions came through fStik and QuotLy. I also won first place at the official Telegram Mini App Contest. Useful proof, though production has taught me more than contests ever could.

I like work with consequences: systems that hold real data, save real hours and still have to behave on a bad Monday morning. That is the thread connecting enterprise AI, open-source experiments and Telegram products at scale.

I write about the work and what I learn at @LyBlog.

How I build

For about two years now I've done most of my coding alongside AI agents. Day to day it's Claude, currently the strongest tool I've tested; GPT and Gemini I keep around for verification and comparison. Working stack: TypeScript on Node.js, React / Next.js on the frontend, Python for ML, Telegram through Telegraf, grammY, TDLib, MTProto and Mini Apps. All on my own boxes: Linux, Docker, Sentry, MongoDB / Postgres / Redis

Get in touch

Have a workflow that wastes a team’s time, an AI system stuck at demo stage, or a product that needs to survive real scale? Tell me what is broken.