📍 Paris, Station F · On site · Full-time · Permanent

Build agents that do mechanical engineering

Faber builds AI agents that carry out mechanical engineering work: creating CAD models, producing technical drawings, inspecting parts, and preparing manufacturing.

Our users work with physical parts. A model that looks convincing can still have the wrong dimensions. A successful tool call can still produce an unusable result. Making these agents useful means giving them the right tools, the right feedback, and a way to check their work.

We’re looking for an AI engineer to build these systems and turn them into products engineers can rely on.

You’ll own substantial parts of the agent and product, make architectural decisions, and take responsibility for how they behave in production.

You’ll work directly with customers to understand their engineering workflows, build a solution they can use, and bring what you learn back into our core agent and product.

What you’ll work on

• Build and improve agent loops: tools, context, execution, feedback, and recovery.
• Turn a customer’s engineering problem into a working product, from the first experiment to a workflow they can use themselves.
• Develop evaluations that show whether a change actually helps. Investigate failures and turn them into better tools, experiments, and checks.
• Build the systems that run agents reliably, including concurrent jobs, traces, and cost tracking.
• Create tools that help you and the team develop faster: inspect runs, compare approaches, debug failures, and understand model behavior.

You’ll work closely with the founders, geometry engineers, and customers. Some improvements will come from a new model or a better prompt; others from changing what the agent can see, do, or verify. Your job is to find out which ones matter.

You might be a good fit if

• You have strong software engineering fundamentals and are comfortable building and debugging in Python.
• You have shipped systems with language models and tools, and can explain where they worked, where they broke, and the engineering decisions behind them.
• You take end-to-end ownership, from understanding the customer’s problem to shipping a solution and making sure it works in practice.
• You show high agency and a bias for action: you turn ambiguous problems into concrete next steps, seek out missing information, unblock yourself, and iterate quickly with feedback from users and experiments.
• You have strong product judgment: you care about usability, speed, and polish, question assumptions, and make pragmatic tradeoffs that help the customer get their work done.