About
For over ten years I have worked where data, software and operations meet — from a Swiss federal apprenticeship in IT, through a degree in Data Science and Economics at Lucerne University of Applied Sciences, to my current work as a Data & DevOps Engineer in Zurich.
What interests me is the part after the prototype. A pipeline that runs in a demo notebook is not a data platform. An agent that impresses in a workshop is not yet an application you can operate. Getting there — governance, testing, cost control, monitoring, a clean handover to a team — is the part I find most interesting.
What I'm building
What I work on
Data Platforms & Lakehouse
Central data platforms on Azure Databricks: ingestion, transformation, modelling, governance. Delta Lake and Unity Catalog, Spark and SQL workloads, custom Python frameworks for metadata management.
AI Applications & Agents in Production
The step from prototype to operable application: evaluation and regression testing for LLM systems, monitoring, cost control, access control. Especially in regulated environments, where “usually works” is not good enough.
DevOps & Delivery
CI/CD for data and ML workloads, infrastructure as code, test and deployment pipelines. Increasingly agentic: coding agents for reviews, tests, migrations and documentation — agentic DevOps as delivery machinery, not as a buzzword. Plus establishing development processes and setting a team up so releases become boring.
Posts
What is an AI factory?
The definition, the six levels from autocomplete to dark factory, and what it takes to move up.
orch: ticket-based work with coding agents
orch: signed approval gates, evidence instead of claims, and the tools I use every day. The product: orch core.
Certifications & Degrees
Experience
Over ten years across data, software and operations — from the data centre through data warehousing to the Databricks platform. The part-time roles from 2022–2025 ran alongside part-time studies at Lucerne University of Applied Sciences.
-
since Sep 2025Data & DevOps Engineer / Consultant — ALTYCA, ZurichCentral data platforms on Azure Databricks, CI/CD automation, migration from Qlik to Power BI, establishing development processes and tooling.
-
Jan–Sep 2025Data Engineer — b.telligentData warehouse solutions and ETL pipelines with Databricks, Snowflake, Azure and Microsoft Fabric.
-
2024–2025CTO & Full Stack Developer — Voicy AG TeilzeitVoice-first web app for organisational self-reflection with AI-based data analysis.
-
2023–2025CTO & Data Scientist — Layer.swiss TeilzeitBuilding Layer.swiss, a spin-off of Oberholzer Digital, for data-driven business solutions in SME marketing.
-
2022–2024Consultant — Oberholzer Digital TeilzeitData-driven strategies and web analytics stacks for clients.
-
2020–2022Application Manager — Screenimage Systems AGImplementation of digital signage software and automation solutions.
-
2013–2020System Engineer — Bechtle SchweizOperating and developing systems and cloud services in the data centre; Swiss federal apprenticeship in IT systems engineering (EFZ).
Technologies
Projects
Personal projects alongside work — where I try out whatever is occupying me technically, and learn fastest. Code and further experiments live on GitHub.
orch core — let coding agents do the work, keep every decision yours
Every piece of agent work is a ticket with three gates: requirements, plan and verdict. A person signs each one with a fingerprint or their phone, so no agent can sign; the verdict signs the exact commit the evidence belongs to. Mission Control, an iPhone app, and AI Factory for work that runs inside limits you signed. Free for noncommercial use (PolyForm Noncommercial 1.0.0). → severin.io/orch-core · GitHub
Fernscout — a self-hosted travel journal
An open-source travel journal you host yourself: the place, the route, the photographs and what it cost, written up each day in your own studio — with your own agent or the built-in assistant as an optional second way in. Family and friends read along in a browser. Everything stays JSON documents and photographs in a folder you own. → fernscout.ch · GitHub
HFT Trading on Polymarket & Hyperliquid
Custom trading infrastructure for high-frequency strategies on Polymarket and Hyperliquid: real-time market data ingestion, order execution under tight latency budgets, risk limits and end-to-end monitoring. A testbed for what a system looks like when it has to run unattended around the clock.
Swiss AI Weeks — Banking Client Simulation
Synthetic banking datasets and personas for testing AI systems for fraud detection and customer analytics under realistic conditions. Scenario engine for configurable customer behaviour. → GitHub
Contract Generator
AI-based platform for producing legally valid contracts — encrypted, with digital signature. For me primarily a study of what an AI application needs in order to actually be operated in a legally sensitive environment: traceable generation, clean templates, auditability.
LoRa IoT — Sailboat Sensor
Self-built low-power sensor for GPS, temperature and humidity, with live tracking over LoRa and a web dashboard.
Contact
Building a data platform, or working on getting AI applications into production? I am always glad to talk shop — no particular occasion needed.
hello@severin.io LinkedIn GitHub
Zurich, Switzerland · Usually replies within a day or two.