Physicist · software engineer · 10+ years in R&D

With AI, building is cheap.Knowing what to build isn’t.

I build and fix software where the numbers have to be right: models on technical and experimental data, AI tools that get checked rather than trusted, and scientific and engineering code that has to keep running for years.

  • Life sciences & diagnostics
  • Instruments & sensors
  • Industrial R&D
  • Physics & simulation

What I do

  1. 01

    Scientific & engineering software

    New builds, refactoring and legacy modernisation (typically Fortran or MATLAB), with tests that pin down the results first.

  2. 02

    Data & machine learning on technical data

    Predictive models, signal processing and imaging, with uncertainty you can report.

  3. 03

    Modelling & simulation

    Physical and statistical models of what lies behind the data, numerically stable, from prototype to production.

  4. 04

    AI applications, verified

    LLM document analysis, data curation and evaluation, built so you can check the output.

Selected work

Results from client projects and my own tools.

Legacy refactoring

Up to 15× faster scientific code

Large legacy scientific codebases refactored and optimized under a strict test regime: faster, easier to maintain, results unchanged.

LLM analysis

800+ SOPs turned into KPIs

A multinational’s internal SOPs read and turned into structured KPIs, so leadership could spot gaps and standardize across teams.

Algorithms

Terabyte-scale data, interactive

Terabyte-scale data made smooth to explore on consumer hardware, for technical operators.

Lab software

Pharmacopoeia-compliant assay software

Replaced Excel: from raw data to the reported result and its uncertainty, readable by lab staff.

Modelling · own product

synfit

Dose-response and drug-synergy fitting with explicit noise models. Python package and web app.

File formats · open source

bioio-vsi

Reads Olympus/Evident microscopy images without Java. Undocumented format reverse-engineered, validated against Bio-Formats.

Rust · open source

wfdisc

Rust library for seismological waveform data, cross-validated against ObsPy.

Image analysis

Physio_Ca

Microscopy video pipeline: from days of manual analysis to under an hour plus a short expert review.

About

Physicist by training: a PhD on neutrino oscillations, then simulation work at the Max Planck Institute for Physics. From there to computational biology at ISTA and biomedical data analysis at the Medical University of Vienna, then to industry as a senior data scientist at Zühlke. Independent since 2024.

What carries over between fields: models that explain rather than only predict, uncertainties stated honestly, and software that keeps working after I leave.

Working on something where data, modeling or simulation matter?

Tell me about it. I reply within a day or two.