BASED IN BREMEN
Data pipelines and AI for the German Mittelstand.
I build the path from the machine to data operations and IT can actually use. For plant, maintenance and IT leads — from Bremen, remote across Germany.
No obligation · reply within 24 h
- Selected projects
- Separate from employment
- Remote in Germany
ABOUT
Usable data first, then AI.
I'm Eugen Schmidt. From mechatronics and maintenance to data engineering — I know the machine and the data path. I build the foundation that reports and AI can hold in production.
By day I work at one of the world's largest automotive groups in Bremen. Freelance I take selected projects — cleanly separated, with agreed times.
In industrial production since 2008, a data engineer since 2021.
Project-based, for selected work. Not a full IT outsourcing.
Based in Bremen — remote with companies across Germany.
SERVICES
What I take on.
No toolkit pitch. A few clear blocks, in the language of operations and IT.
Data pipelines
From sensor or system to a table someone actually uses — so reports and AI don't fail on bad data.
Machine data
From the controller to a number you can trust. Siemens, MQTT, the path across the hall — in plain language.
Read moreAI in your own house
When data must not leave the site: models and pipelines in the EU or on-prem.
Read moreDigitalisation for SMEs
Prepare and implement the IT, including when BAFA funding is in the room. No rulings, no tax advice.
Read moreDATA STAYS HERE
AI you can run in Germany.
For mid-sized firms this is often the prerequisite, not an extra: operations data does not automatically belong in any cloud.
EU residency
Data and models in EU data centres — no default path to a US cloud.
On-prem when needed
Where the data requires it, the path runs on your network. Nothing has to leave the site.
GDPR from the start
Who processes what, where it lives, which notes you need — part of delivery, not an afterthought.
In plain terms: your machine data does not have to leave the building — and the processing notes come with it.
More on on-prem and EU AISELECTED WORK
From the factory floor, not a slide deck.
Anonymized glimpses from my work as a data engineer in automotive. I set up the same kind of system at small scale too: one machine, one server, a scoped data check.
Manufacturing
1M+
machine signals processed per month in real time
Monitoring from sensor to readout
PLC and sensor data carried over MQTT and analysed in time — with a warning before a process stops.
Planned maintenance instead of firefighting — downtime becomes avoidable.
- Siemens S7
- MQTT
- Databricks
Maintenance
Seconds
to an answer instead of searching manuals
Assistant for machines and manuals
Fault questions answered from manuals and live machine data — on your network, not in an arbitrary cloud.
The shift finds faults itself, without waiting for an expert.
- RAG
- Vector Search
- on-prem LLM
Data platform
10×
faster onboarding of a new machine (weeks → hours)
One architecture, many machines
The monitoring path as a reusable blueprint — a new machine is connected faster than a rebuild.
Every new machine is connected faster and cheaper.
- Databricks
- PySpark
- FastAPI
Team reporting
0 SQL
answers in plain language, in the tool they already use
Ask production data in plain language
Teams ask for KPIs in the tool they already use, in plain language — without waiting on the reporting queue.
Answers without waiting on the BI team.
- Power BI
- Databricks
- LLM-Serving
From architecture to delivery, with the teams on site — anonymized for confidentiality. No client names, no internal system names.
Sounds like your topic? Let's talkTOOLS WHEN THEY MATTER
Not a promise dressed as a toolkit — these are what I actually use.
Data paths
- Databricks
- PySpark
- Delta Lake
- Unity Catalog
Shop floor
- Siemens S7
- MQTT
- Node-RED
- Event Hub
AI
- Vector Search
- RAG
- LangGraph
- LLM-Serving
Applications
- FastAPI
- React
- Recharts
- Power BI
Operations
- Docker
- Hetzner
- CI/CD
HOW I WORK
In four clear steps.
Intro call
No obligation: we clarify the goal, your data, and whether the work is worth it.
Analysis & proposal
I look at the status quo and propose a concrete, bounded plan.
Delivery in stages
Short steps, a visible state — not a months-long black box.
Handover & docs
Documentation so your team can continue. Ongoing ops only if we agree.
Typical start: a scoped data check or pilot, often within a few weeks — then you decide how to continue.
Request the intro callCONTACT
Describe the situation.
A few sentences are enough. I'll reply within 24 hours.
No obligation · reply within 24 h · start with a scoped check.