About me
Stefan Koch
Architect · Trainer · Databricks Champion
I build data platforms you can still understand two years from now — and teach others to do the same.
In short
I started out as a machine mechanic. Then came fifteen years with the Zug cantonal police — traffic and response policing, regional police, criminal investigation. For the last eight of those I worked in digital forensics. So I have been working with data since 2010; back then it just wasn't quarterly figures, it was evidence.
That period still shapes how I build systems. In forensics the best analysis is worthless if you cannot trace every step of it: where did this value come from, who touched it and when, does the chain hold up in court? Those are the same questions I put to data pipelines today. Traceability is not a feature you bolt on later.
In 2018 I moved into data consulting — through Trivadis, Accenture, b.telligent and ALTYCA. The same pattern kept showing up: the platform grows faster than anyone's understanding of it. Eventually nobody dares to change anything.
Alongside that I have been teaching at Lucerne University of Applied Sciences since 2020, where I direct two CAS programmes on business intelligence and data engineering. Teaching changed my consulting work more than any single project did: once you have to explain something to twenty people well enough that it lands on a Saturday afternoon, you start building different systems. Simpler ones.
On 1 September 2026 I founded Nitako GmbH to offer exactly that: building data platforms, enabling teams, and doing both in a way that keeps running without me.
Career
-
since 09/2026 · Rotkreuz ZG
Founder & Data Platform Architect
Nitako GmbH
-
09/2025 – 08/2026 · Basel
Principal Consultant
ALTYCA
-
since 10/2020 · Rotkreuz ZG
Instructor & Director of Studies
HSLU Hochschule Luzern
-
01/2024 – 09/2025 · Basel
Senior → Principal Consultant
b.telligent
-
07/2023 – 12/2023 · Zürich
Senior Consultant Data Engineering
Accenture
-
10/2018 – 07/2023 · Zürich
Consultant → Senior Consultant Business Intelligence
Trivadis
-
2003 – 2018
Police officer – latterly digital forensics examiner
Zuger Polizei
Traffic and response policing, regional police, criminal investigation · including 8 years in digital forensics
-
before that
Apprenticeship and first years on the job
Maschinenmechaniker
Full career history on LinkedIn.
Certifications (selection)
-
Databricks Solutions Architect Champion
Databricks Partner Program
-
Microsoft Certified Trainer
Microsoft · renewed annually, current for 2026
-
Azure Databricks Data Engineer Associate
Microsoft
How I work
Understandable beats clever
A platform only I can operate is not a solution, it is a liability. I document, hand over and make myself redundant on purpose.
Start small
A single piece of pipeline moving real data next week beats a target architecture that is still a diagram six months in.
The two-year test
I build nothing I would not want to operate myself two years later. That rules out a surprising number of tools — and saves a surprising amount of money.
What I work with (selection)
- Databricks
- Azure
- Azure DevOps
- GitHub
- Cloudflare
- Hetzner
- OpenTofu / Terraform
- Docker
- Python
- SQL
Tools are a means to an end — the choice follows from what your team can operate afterwards.
I write about it
On my personal blog I document what I am currently working on — usually with code you can rebuild yourself.
All posts on stefanko.chCross-published on Medium.
Learn Databricks SQL - Part 2 - Where and how to run it
The same two SQL statements, sent to a Databricks SQL warehouse in every way I could test: as a job with a SQL file, as a notebook, with the CLI, the REST API and from Python. Where a script of several statements runs, where it fails, and where the session is lost between two calls.
Learn Databricks SQL - Part 1 - Why, and the data we'll use
The start of a series that goes through the SQL of Databricks statement by statement, every one run on the same small tables. This part: why, what you need, and the sample data – a fictional bike rental in Lucerne with six tables and a few rows that are wrong on purpose.
Live Flight Data Step by Step - Part 17 - Looking back, and the code
The last part of the series: what 22 hours of crowdsourced aircraft positions could answer, what worked, what didn't, what I'd do differently – what's left to analyse, and the code as a public repository.
Sounds like your project?
Tell me briefly where things are stuck. I usually reply within 24 hours.