Build the runtime the frontier labs kept to themselves.

Zibra Labs is building the post-training and inference infrastructure that today exists only inside the frontier labs. Our vision is to let everyone post-train and serve open weight models at a fraction of the cost, with higher capability.

We are a small team in San Francisco, working in person. The problems are the deep kind: schedulers, memory, and networking on specialized hardware, at a scale where the details decide whether the thing works.

Open Roles

Applying

Send a note about what you have built and what you want to work on. Skip the cover letter. A repo, a design doc, or a story about a system you took from broken to boring tells us more than a résumé does.

Nothing here quite fits, but you think you should be on this team? Write anyway and say why.

Our address: careers (at) zibralabs (dot) ai