---

My Story

I started writing code in 2007 — not because I studied computer science, but because I couldn’t stop. My formal education was in metallurgy and pharmacy, neither of which I completed. What I did complete was roughly 18 years of building real-world web applications, debugging production systems at 2am, and leading teams through rewrites, migrations, and launches.

I didn’t follow the traditional path into tech. I learned by building — small projects, then bigger ones, then projects with real users depending on them. That’s still how I work.

---

What I Do

I’m a Web Development Team Lead at FireBit LTD, where I oversee full-stack development across multiple products. My day-to-day spans architecture decisions, code reviews, mentoring developers, and hands-on engineering when the problem is interesting enough.

On the side, I run kalyna.pro — this blog. I started it to document what I’m learning about AI and LLM integrations. Writing about something forces you to actually understand it.

---

Tech Stack

My core stack has evolved over 18 years, but the constants are:

Lately I’ve been spending most of my learning time on: Python, the Anthropic Claude API, LangChain, vector databases (ChromaDB, Pinecone), and building AI agents. That’s what you’ll find most of on this blog.

---

This Blog

kalyna.pro is an AI Dev blog aimed at working developers — people who already know how to code and want to understand how to work with LLMs effectively.

I write about:

• Claude API and Anthropic’s ecosystem — practical tutorials, not marketing
• Building AI agents and multi-agent pipelines in Python
• RAG (Retrieval-Augmented Generation) — real implementations, not toy examples
• Tools I actually use: Cursor AI, n8n, LangChain, vector databases
• Cost optimization and production patterns for LLM applications

Everything is based on actual usage. If it’s on this blog, I’ve built it.

---

Contact

Best way to reach me: [email protected]

I’m open to consulting on AI integrations, code reviews, and architecture questions for teams adding LLMs to their stack.