Unstructured in.
Trusted data out.

I've spent eight years moving across the stack — web development, then QA, then data engineering, and now AI. That path wasn't an accident: each stop taught me a habit the next one needed. QA made me distrust every record until it's validated. Warehousing taught me that pipelines fail at 3 a.m., so they'd better log well. LLM work taught me that the model is only as good as the context you engineer around it.

Today I work on LLM-powered document processing — the unglamorous, high-value problem of turning contracts, invoices and scanned forms into data a business can actually trust. On the side, I take freelance projects and study financial engineering, because markets are just another messy data source waiting for structure.

Since 2022 I've also taught programming to kids aged 6 to 18. Four years of explaining recursion to ten-year-olds is the best communication training an engineer can get — if you can't make it clear to them, you don't understand it well enough yourself.

Off the clock I post about data and AI as @ajdin_thinks_data — the handle is honest, it's how my brain works. I speak Bosnian and English fluently, get by in Slovene, and I'm slowly convincing German to cooperate.