Niklas Savonheimo
Data & backend · Helsinki
About
I'm an IT engineering student in Helsinki. My coursework is done and my thesis on entity resolution is in progress. I build data pipelines and small tools, mostly things that run on a schedule and keep running without me.
Professio, April–September 2026. Five months of customer data work in a CRM, supporting sales and marketing. I ran four lead-generation projects, built and delivered 170+ segmented contact lists, and added or enriched tens of thousands of records. New job titles kept landing in one oversized category. I worked out what distinctions the incoming requests actually needed and proposed five new categories. They went in. Contact verification was in Finnish, with Swedish and English alongside.
firm-fiend. Company registries from Finland, Sweden and Norway pulled into one shared model. Each country ships its data in a different shape. Python, dbt, DuckDB, Docker Compose.
match-jobtitles. An offline pipeline that matches English job titles to a fixed Finnish category list using local embedding models. Validated on 5,400 records against about 380 categories, at about 92% coverage.
fuel-dash. Scrapes crowdsourced fuel prices twice a day, builds up the history in SQLite and publishes a dashboard. It has been collecting since July 2026. If a run fails, the next one picks it up.
I like data work especially. There's something satisfying about getting a result out of it, even if it's just an optimization that saved half a cent.
Open to data, backend and ML work.
Python · Git · Docker · Excel · Power BI · pandas · scikit-learn · SQL · dbt · DuckDB
Selected work
Recent picks — repo and live links below each.
For the rest, see all projects →
Eduskunta Analysis
Finnish parliamentary voting records from 1996 to today. Maps which MPs actually vote alike, compared with the official party split.
Writing
Things on my mind, written out.