Hi, I'm Jan. I previously worked as a software engineer and am currently transitioning into AI safety research. I most recently worked as a Software Engineer at DeepL, tried to build a startup at Entrepreneur First and worked at Meta before that. Feel free to write me if you find any of my work useful or interesting.
Building skills toward AI safety research.
- pytorch_exercises — ML model training experiments, becoming more proficient with PyTorch.
- numpy_exercises — Worked through the first 70 of Nicolas Rougier's 100 numpy exercises.
- ai_control_hackathon_monitor_finetuning — Fine-tuning Qwen-3.5-9B to act as a stronger monitor over larger untrusted models. Apart Research, March 2026.
- news-agent-anthropic-hackathon — Daily digest recommendation agent built on Claude. London Anthropic Hackathon.
- advent_of_code — My Advent of Code solutions across several years and languages; in recent years I used it to get better at Rust.
- fullstack_project_template — Vibe-coded fullstack template: Rust backend, Svelte frontend, built-in SQLite, user auth, JWTs.
- elo_rating_implementation — Small exploration of the ELO rating calculation in Python.
Official client libraries and API things I worked on as part of the DeepL API track.
- DeepLcom/deepl-python — Python SDK
- DeepLcom/deepl-java — Java SDK
- DeepLcom/deepl-dotnet — .NET SDK
- DeepLcom/deepl-php — PHP SDK
- DeepLcom/deepl-ruby — Ruby SDK
- DeepLcom/deepl-node — Node.js SDK
- DeepLcom/deepl-mock — Mock server for the DeepL API
- DeepLcom/api-docs — DeepL API documentation
- DeepLcom/openapi — DeepL API OpenAPI specification