Tooling · Benchmark
Measuring what a local LLM actually costs on a GPU-less machine.
- Python
- Typer
- Pydantic
- Docker
- GitHub Actions
Work
Six projects, each with its own page: the original problem, what I actually built, and the technical points that required a decision. The code is public — check it rather than take my word.
Measuring what a local LLM actually costs on a GPU-less machine.
Running a multi-agent system at home, without the cloud, and keeping control of the data.
A personal assistant collecting intimate data must first explain what it collects.
Portfolio documentation always happens at the end — so never. Let us automate it.
Unstructured research produces unverifiable conclusions. The method matters more than the tool.
This showcase: static, bilingual, no CMS and no database.