I build small, useful things in Python — and I can explain them to people who don't write code.
Four years of Python, nine of SQL across PostgreSQL, Oracle and SQL Server. Most of my work has been in places where the data was real and messy: a Swiss tram network, a waste-management utility, a health-insurance system, an avalanche model. I care about tools that someone actually opens and uses.
Python (Flask, pandas) · SQL (PostgreSQL, Oracle, SQL Server) · Git · Docker · REST APIs · GIS (QGIS, spatial data) · Jupyter
I use large language models daily as a working tool, with structured prompts — role, context, constraints, task, examples, output format. One concrete example: I used a model to port my avalanche algorithm from R to Rust, a language I don't otherwise write. That is roughly what LLMs are good for — and it is also why I check every output before it goes anywhere. I was programming in Python and R before they existed, which is where that habit comes from.
Insurek — swiss-deductible-chooser
A web app in four languages that works out the cheapest Swiss health-insurance deductible. Flask, Docker, no data stored. The whole answer turns out to be one line — below roughly CHF 2,000 of yearly expenses, take the CHF 2,500 deductible; above it, take CHF 300 — and the app shows you why, for your own numbers.
network-availability
Written during my time at Basler Verkehrs-Betriebe: how much of Basel's tram network is slowed down or blocked by construction works, at any point in the year. Python, SQL and geodata in, matplotlib reports out — the kind that operations people read before deciding.
Avalanche-release-improved
An algorithm that finds potential avalanche release areas from terrain data. It began as my master's thesis, became an article in a technical journal, and I've kept improving it since.
user-agent-parser
Descriptive statistics on who actually visits Skitourenguru.ch — small, but it answered a question its owner had and couldn't answer.
I've published three books and I write in Italian and English. I also did the scientific and technical review of an avalanche safety handbook for an Italian publisher, and I work three four languages.
This is not a hobby I keep at a distance from the code. Writing three books taught me that good writing comes out of good thinking, and that if a reader doesn't understand you, the problem is yours. That's the same discipline a good README, a clear error message or an honest interface needs. On projects where the people building and the people using speak past each other, I'm usually the one translating.
Right now I'm finishing a book called Shel and building Ophanim, a set of questions that helps someone work out what their own work in life might be — designed to be understood by people who left school at fourteen, which is a much harder design problem than it sounds.
Open to Python and data work, remote or in north-eastern Italy.