I take over, refactor, and automate grown Excel, VBA, Access, reporting, Power Query, Python, and AI workflows. Until a second developer understands them in a week.
What started as an Excel file now carries half a business process. What was meant to be a Python script from the ex-intern is now production. I clean up, document, make it handover-ready.
/ Stack
/ Stance
I do not build slick prototypes for LinkedIn. I build systems that still run in two years, that the team can keep alive without me, and that do not crumble at the next update.
Excel, VBA, Access, SQL, Power Query, Python, local AI, classic reporting pipelines. The tools are not the interesting part. The interesting part is taking a grown-over process and making it predictable again.
01 // Focus
Four constellations I get called in for most often. Each with a concrete before-and-after, no stock imagery.
Same exports, same pivot, same five hours. We structure import, logic, and output once, usually Power Query plus a small VBA layer.
Nested formulas, broken references, macros from 2014. Refactor into clear zones, typed ranges, documented steps. Nobody has to guess what happens.
Grown-over Access apps where business logic, data, and UI are one big soup. Carefully untangle, document, and optionally migrate to SQL Server.
n8n pipelines, local models, embedding stores, MCP servers, document analysis. Pragmatic architecture instead of cloud lock-in for data that cannot leave.
02 // Architecture · sample
A schematic, not a mockup. What comes out of the discovery workshop tends to look about like this. The clarity is the point: who does what when, where the data comes from, where the responsibility sits.
03 // Disciplines
A compact index so the scope is clear at a glance. Each discipline has its own service page with problem, approach, and typical deliverables.
04 // Engagement
Seven formats I offer most often. Each has a clear scope, a clear start, and a clear point where you decide on your own terms whether more makes sense.
All amounts net. Small fixes under one hour are not billed. Larger pieces can move to a fixed-price package once the scope is clear.
Short get-to-know call. You outline the topic, I give an honest first read on whether the project makes sense at all. Directly with the engineer, no sales.
Longer session with written assessment: data flow, risks, three realistic options with sequence. You can think it through internally, with or without me.
Review of an existing Excel, Access, or reporting estate. Risks named, concrete sequence for change proposed.
Where AI realistically helps in your stack, where it does not, what GDPR and the EU AI Act mean for it, which architecture holds.
The actual work: code, data model, pipelines, refactoring, migration. Iterative, documented, handover-ready.
Agreed monthly hours for maintenance, adjustments, urgent questions. Response within two business days.
In-house, remote, or hybrid. Excel, VBA, SQL, Power Query, or an AI block tailored to your team.
Anonymized, roles preserved. No logo wall here, because trust does not come from logos. It comes from what still works after the project.
Finally someone understood why our monthly reporting was broken in the first place, before trying to fix it. Everyone else only ever knew the symptoms.
No AI coach theatre. An honest assessment of whether our use cases were ready for local inference, and a sensible sequence plan.
For the first time we did not have to click our weekly reporting back together by hand. Five hours of clicking became one button. Sounds banal. It was a small liberation for the team.
Understood our data flow first, then our tables, then the actual process. Only after that did code enter the picture. Exactly the sequence we were not used to.
Our Access app from 2009 was still running, but nobody dared touch it. Three weeks later the code was documented, the backend mirrored to SQL Server, and the app ran faster than before. Without a single user noticing.
An Excel file that had been growing for eight years and nobody really understood. Refactored in two weeks, clear zones, documented logic. What surprised us: he also told us which parts to leave manual on purpose, because automation would not pay off there.
We wanted to throw ChatGPT at our contract base. What we got was a local RAG setup with citations, GDPR-compliant, on-prem. Exactly what our data protection officer had been saying we would need all along.
/ FAQ
Not an SEO FAQ. This is what people genuinely want to know before the first conversation. Answers are as short as possible, and as long as necessary.
/ Engagement
/ Existing systems
06 // A short note is enough
Send a short email about what still runs manually. You get an honest read back. No sales funnel, no discovery-call sequence.