I’m Sebastian, a data engineer in Wrocław, Poland. I build the unglamorous middle layer of data work: pipelines that run on a schedule, models that are tested, and dashboards people actually open on a Monday morning.
The short version of how I got here
I studied accounting and finance, then business analytics — so I came into data from the reporting side rather than the computer-science side. That turned out to be useful. At Fortum I spent a year and a bit in the sales accounts department writing SQL, building Power BI reports and running ad-hoc analyses, which is where you learn what “the numbers don’t match” really means in practice.
Since then I’ve moved one layer down the stack: from answering questions with data to building the thing that answers them without me. Dagster for orchestration, dbt for the transformations, Python for everything that doesn’t fit either.
What I work on
- bike-data-flow — a Dagster pipeline for Wrocław’s bike-sharing (WRM) station data, with dbt models, an Evidence dashboard and Leaflet maps built on an OSM basemap. Live: seszele64.github.io/bike-data-flow.
- llm-analysis-pipeline — review sentiment analysis on AWS: a DynamoDB stream triggers a Lambda, which calls an OpenAI-compatible endpoint and writes the result back. Boring architecture on purpose.
- docer-downloader — a JavaScript document downloader with a few hundred stars on GitHub, and still the project that brings the most traffic.
- dictator and a pile of MCP servers — Whisper dictation for i3wm, plus tooling that gives coding agents scraping and browser abilities.
What goes in the archive
Notes from building the above. Pipeline design trade-offs, scraping that had to survive a redesign, LLM plumbing that needed more retries than expected, and the occasional static-site post when the site itself is the interesting part.
What this is not
No “top 10 tools you must learn”, no rewritten documentation, no posts I wouldn’t send to a colleague. If a post here is short, it’s because the answer was short.
If you want to talk about pipelines, Dagster, or why your dashboard is slow, write to me at github@harhara.eu.