Fabio Steyer · Berlin
About
I am a physics master's student at TU Berlin and have worked since 2025 where data analysis meets process automation. What draws me to it is the decision before the code: which step needs a script, which needs a language model, and which is better left alone.
Skills and evidence
Each row states what it rests on. "Learning" means present, but not backed by a project.
| Skill | Evidence | Level |
|---|---|---|
| Python | Production pipelines at Cavator; all five Python projects under Projects | publicly verifiable |
| Document parsing | invoice-quote-reconciliation, measured against a ground truth | publicly verifiable |
| Concurrency and data integrity | concurrent-file-writes, measured with and without coordination | publicly verifiable |
| Test discipline | 104 tests across five Python projects, including multi-process runs | publicly verifiable |
| Scientific data analysis | B.Sc. thesis, spectroscopy and model fitting | publicly verifiable |
| OCR | Around 3,000 invoices and quotes at Cavator, not public | used in practice |
| LLM integration (Claude, OpenAI) | Built into my own pipelines at work; the review chain in prose-check-chain is built, recorded model answers are still missing | used in practice |
| Git | Six repositories with a readable history | used in practice |
| SQL | Personal exercises, still building | learning |
On AI assistance
I work with AI tools, including while building this site. The technical decisions, the choice of content and its verification are mine.