DATA PIPELINES
From the signal to a table someone trusts
AI and reports rarely fail on the model. They fail on the path before it: the signal is there, the table is not — or nobody owns the number. I build that path, from the sensor or system to a state operations and IT can actually work with.
Who this is for
IT and operations in mid-sized firms who notice: the numbers sit in five systems, the shift needs one, and nobody can deliver it without hand work. Not for a platform you are meant to buy and fill yourselves — and not for a team that wants to hand over all of IT.
What a pipeline means here
A source, a target, someone who needs the number. In between, quality and freshness: does the value arrive complete, and on time? At the end, a table or an interface someone opens — not a graveyard of jobs. The source can be the hall, a line-of-business system, or both. Where the data may go — on site, EU, a chosen cloud — we settle first. The pipeline follows the rule, not the other way around.
PRACTICE · ANONYMISED
Monitoring from sensor to readout
- Situation
PLC and sensor data from many machines had to be analysed in time — with a warning before a process stops. Until then, connecting each new machine was its own build that took weeks.
- Built
One path from PLCs and sensors over MQTT to the analysis, as a reusable blueprint: one architecture, many machines. A new machine is connected instead of rebuilt.
- Result
More than a million machine signals per month processed in real time; onboarding a new machine takes hours instead of weeks. Planned maintenance instead of firefighting.
What I do not do
No 24/7 platform as a subscription. No replacement for your IT. No promise that every legacy system is 'connected' in one sprint. No tool list sold as the work: if the source is the machine, the first look belongs on the machine data — not on a slide about orchestrators.
How we start
A few sentences: which source, who needs the number, may the data leave the building. Then a bounded first path you can check — completeness and freshness visible, not only 'the job is green'. Handover so your team knows the path; operations only if agreed.
Questions plant managers ask
Is this the same as machine data?
Machine data is often the source. The pipeline is the path to the table someone uses. They hang together; the start sits where your gap is.
Which tools do you use?
The ones that fit the source and the target — not a fixed platform you have to buy. The tools are a means. The delivery is a path someone trusts.
Do you run the pipeline afterwards?
Only if we agree. The default is handover and documentation so your team can keep going. No quiet wrapping of operations.
Does this work remotely?
Yes. The office is in Bremen; the projects are remote in Germany. On site when the hall or the server room decides it.