NiLabs
Join us
We are looking for people who want to work on measurement infrastructure in public. Everything you write is open source under Apache-2.0, with your name on it.
Before you apply
NiLabs is a volunteer open-source community, not a company. These are not salaried positions and we are not offering equity. The capital we are raising covers compute, not payroll. People join because they want the work and the public track record that comes with it.
What you would work on
- Inference correctness — writing conformance tests, adding engine adapters, and chasing down why two engines disagree. Python, no GPU needed
- Training and evaluation — the ladder pipeline: data verification, sandboxed judging, and squeezing runs onto free-tier GPUs
- Upstream contributions — patches to vLLM, llama.cpp and NVIDIA OpenShell, which is where most of our work actually lands
What we look for
One thing, mostly: evidence that you can find a bug nobody else noticed and explain it clearly. A merged pull request, a good issue report, a writeup of something you debugged. We care much more about that than about your background or how long you have been doing this.
You do not need experience with inference internals. Several of the most useful reports we have seen came from people who hit a problem in an application and dug in.
How to apply
Membership is by application. Submit the form with your résumé and a link to something you have built or fixed. No cover letter.
- Your résumé as a PDF
- One link: a merged pull request, an issue you filed, a repository, or a writeup
- Two sentences on which of the three areas above interests you and why
That third item matters more than the first. We read the link before the résumé.
Questions
Email hello@nilabs.dev. Ask before you apply if you are unsure whether your background fits; the answer is usually yes, and it takes us a minute to say so.
Or skip the form
Open a pull request on conform instead. A good divergence report tells us more than any application will, and we will come to you.