Go, Rust & Python 路 Next.js
I build financial research tools, and I fix things in other people's open-source projects. It's mostly the same work either way: API integrations, data processing, and calculations where you can hold the result against a written definition and see whether it holds.
I trade on eToro as a Popular Investor. That's how I got into this, really. I wanted better tools for my own research, and that pulled me into broker APIs, indicator definitions, and the question of what a performance number is actually worth.
Available for open-source contribution work. Debugging existing codebases and building data integrations is where I'm most useful. Portfolio 路 Contact
Pull requests that other maintainers took into their projects:
| Project | What it fixed | Language |
|---|---|---|
| bevy_replicon #764 | Seven places read a size out of a network message and handed it to a cursor that panics when the message is shorter than the size claims. One truncated packet took the client down. | Rust |
| ezBookkeeping #677 | Monthly schedules counted from the end of the month ignored the template's time zone. | Go |
| ezBookkeeping #678 | Edit-scope tests built their boundaries outside the time zone they were testing, so they went red depending on where you ran them. | Go |
| ifc-lite #4900 | The parser pointed its worker URL at a file the package doesn't ship. Now it points at the one it does. | TypeScript |
The repositories are private. What each one does:
Onestein 路 Rust
A messaging protocol stack and library for networks that are slow, intermittent, or simply not there. Protocol logic stays separate from application storage and transport. Experimental, and nobody outside has audited it.
mesura 路 Python
A test bench for trading-strategy claims. You hand it a reported number, it
tells you what the sample behind it can support. Published on PyPI, so
pip install mesura gets you the whole thing. Its browser half, mesura-web,
redoes the trade-record part in TypeScript with every figure checked against
the Python one, and it runs on the portfolio under
/demos/track-record.
factorbase 路 Python, pandas
A catalogue of stock-screening factors. Each one comes with its formula written
out, what it needs as input, and a tested reference implementation. Choices like
smoothing and initialization are documented, which means you can reproduce a
result instead of guessing how it was meant. pip install factorbase.
portfolio 路 TypeScript, Next.js
The source of ludwigjmarx.dev. Every claim on the page gets checked against the GitHub API while the page renders, and comes back as verified, disputed, or unverified with the reason it couldn't be checked. If an API call is refused, the page prints why. It doesn't let a failure pass as a zero.
A bug fix starts with a failure I can reproduce. Then a regression test that is red before the change and green after it, and I run both rather than claim them. The diff stays on the problem. Commands, output, and why I took this route and not the obvious one go into the PR, so a reviewer can check the work instead of taking my word for it.