Join the team
Build the tools
engineers reach for.
We are a small team in Paris teaching machines to read and write geometry. Applications are open the year round, whether or not one of the roles below is yours — send your CV and we will read it. If you are looking for the product instead, try the beta; if your company wants to build something with us, work with us.
Open roles
Four ways in.
Founding Mechanical EngineerDirect the product, work with our customers, build the evaluation pipelines.Paris · On site · Permanent
You set what Faber builds next, and you work directly with the engineers using it. You also own how the agents are measured: the real parts they run on, the checks they are scored against, and the call on a tolerance or a wall thickness that no automated check makes for us.
It needs years in industry — parts you took through process selection, tolerancing and inspection, with something at stake if you were wrong. That experience is the contribution, and there is no way to acquire it here.
Apply with the form at the top of this page, and name the role in the note.
Founding AI Engineer · Agent harness & infrastructureBuild the harness the agents run in, the infrastructure under it, and the evals.Paris · On site · Permanent
You own everything between the model and the result: how a part is represented on the way in, what the agent can call, what happens when an operation fails, how a run is replayed when it goes wrong. The infrastructure that runs hundreds of attempts on real files is yours, and so are the evals that say whether any of it improved.
No CAD experience needed. What is needed is agentic systems you have built that had to work rather than demo.
Apply with the form at the top of this page, and name the role in the note.
Founding AI Engineer · Post-trainingPost-train multimodal models to reason about physical and geometric problems.Paris · On site · Permanent
You own the post-training: the reinforcement learning runs, the reward design that decides what the model actually learns, and the data both rest on. The subject suits it — a part satisfies a constraint or it does not, so the reward can come from a real check rather than a preference model.
You should have post-trained something and seen a number move. Depth in reinforcement learning, multimodal training or data work all count.
Apply with the form at the top of this page, and name the role in the note.
AI InternshipData/engineering and research positions.Paris · On site · 4 to 6 months
You take one thing and carry it to the end. On the data and engineering side that is an evaluation set that does not exist yet, or the infrastructure to run it at scale; on the research side, a modality the agents cannot handle, or a benchmark that tells us something we did not know.
A background in AI is required.
Apply with the form at the top of this page, and name the role in the note.