I understand the bookkeeping, the spreadsheet and the process behind it — and I build the tool that replaces it.
Before the bootcamp I studied business, helped build sales and operations at an AI start-up, and today I sit on the board of a foundation, where I am responsible for its finances. So I know what an open-items list looks like, why your colleague's spreadsheet exists in three versions, and what the accountant needs at month end. Since 2022 I have been building the software for exactly that: small, robust tools that follow the process instead of bending it.
Four things small businesses almost always lack – and that can be built in weeks once you understand the process behind them.
When a spreadsheet has turned into a database, it is time for a program. I build desktop and web applications with a real database, a clear interface and import/export – installed on your machines, not sold as a subscription.
Numbers that are current every morning without anyone copying them together. From accounting exports, sales lists or databases comes a dashboard that answers the three questions you actually ask.
The DATEV export, the supplier's list, the spreadsheet from the online shop: I connect the sources, clean the data and let the routine run on its own – nightly, daily, no manual work.
Machine learning only pays off when the data supports it and the question is clear. If so, I build the model myself – classification, image recognition, forecasts. If not, I say so up front. Language models go where they take over text and routine work.
One paid client project, a model that reads guitar recordings, and a few things I built because I needed them myself. Source code wherever it is public.
2026
Replacing a 1983 DOS system for a long-established Munich company
The legacy system: BASIC from 1983, over 500 source files, data in hand-managed binary files. Reconstructed byte by byte, rebuilt as a Python desktop app on SQLite, accounting logic 1:1 after the original. In production on several machines.
2021 · 2026
Guitar tablature from audio, via CNN
Keras 3 over CQT features, 6 strings × 21 classes. Modernised in 2026: Python 3.12, uv, CLI, Streamlit app, Docker, pytest, GitHub Actions. Calibrating the silence class raises the F-score roughly threefold — with no retraining.
2026
Deterministic, replayable rules engine for a trading card game
A pure reducer with seeded RNG, card effects as a data DSL, around 200 cards, 57 tests, and a browser match against an AI.
2025 · 2026
Fullstack application with image recognition
Upload a photo, identify the card by OCR, manage the collection, export as CSV. React 18 and Vite on the front, Node/Express and Sharp behind, wired to the Cardmarket API.
2026
Data pipeline and dashboard
CSV to JSON into an HTML dashboard running as a pywebview app, daily price snapshots over an API, an online-learning model scored by Brier score, orchestrated with n8n.
Two tracks that cross in 2021: first business, sales and administration, then data science, software and teaching. The crossing is what I offer.
2014–2019
Munich Business School
Focus on finance and corporate governance
2017–2018
Chatbot Consulting · Chatbot News Daily
Advising on chatbots in customer service; market analyses, articles
2017–2019
e-bot7 — AI for Customer Service, Munich
Introduced a sales automation tool, prepared data for NLP training
2021–2021
Le Wagon, Munich
Python, SQL/BigQuery, Scikit-learn, TensorFlow/Keras, GCP, MLflow
2022–now
Self-employed, Munich
Internal tools, dashboards, data pipelines and automation
2023–2025
Le Wagon, Amsterdam
Python, SQL and machine learning; mentoring and code review
2025–now
Biehler von Dorrer Foundation, Munich
Finances, tax filings, liaison with the tax firm
Data science bootcamp
Le Wagon · 2021
B.A. International Business
Munich Business School · 2019
A-Levels
Bede's Senior School, UK
Codecademy — BI Data Analyst Career Path
2024
DataCamp — Tableau
2023
DATEV — practice course
BVH — stock exchange certificate
Python
Pandas · NumPy · Scikit-learn · TensorFlow/Keras · Streamlit · PySide6 · pytest
SQL
BigQuery · SQLite
JavaScript / Node
React · Express
Operations & automation
Docker · uv · Git/GitHub Actions · n8n
Business
DATEV · Excel · Buchhaltung/Steuern · Business Development · Projektmanagement · Lehre
Languages
German (native) · English (fluent — secondary school and degree in English)
We look at the process and the spreadsheet together.
In weeks, not months. Weekly demos.
Runs on your machines, documented, with support.
Send me two sentences: how the process runs today, and what is annoying about it. You get an honest assessment of whether and how it can be built – no slide deck, no sales pitch.