Founding AI Engineer

You would be the founding engineer on the core of the product: the models and agent systems that translate written requirements into PCB designs the tools agree are correct. Off-the-shelf models do not do this well enough on their own, so the work runs from the current literature through training runs, reinforcement learning on constrained design problems, fine-tuning and custom architectures. It then runs into the half that decides whether any of it ships: evaluation harnesses, inference cost, tool-use reliability and the backend that serves it. You would make those calls with the founder and own them from first experiment to production, most of it in the open alongside the developer and hardware community already using the tool. The bar is research-level and the setting is a production team, so a result that does not survive an eval does not ship.

What you would own

• Build AI agents that plan, generate, modify and verify PCB designs.
• Train, fine-tune and evaluate models for the parts of the design problem where a general model is not good enough.
• Apply reinforcement learning to constrained design, placement and routing problems.
• Design custom architectures, including transformer encoders over schematics, netlists and layout.
• Develop the tool-use, context, orchestration and evaluation systems that decide what ships.
• Improve schematic generation, PCB layout, component selection and datasheet understanding.
• Integrate LLMs with KiCad, deterministic verification tools and open-source EDA infrastructure.
• Design production-quality backend systems and developer-facing interfaces.
• Improve system reliability, performance, observability and test coverage.
• Own projects from first experiment through to production deployment.
• Contribute to technical documentation and open-source engineering practices.
• Help establish copperhead’s engineering culture and development processes.

Hard requirements

• Working knowledge of machine learning deep enough to derive, implement and debug the methods you use rather than call them.
• You have trained and fine-tuned transformer models, and you understand their internals well enough to change them.
• You have implemented reinforcement learning algorithms yourself and applied them to constrained or combinatorial problems.
• You have designed custom architectures, including encoders over structured or graph-shaped data, rather than only fine-tuning what already exists.
• You read current AI research, can tell a real result from a benchmark artefact and can implement a paper from its description.
• Experience designing evaluations, benchmarks and reliability metrics for systems whose output has to be correct.
• Strong software engineering in Python, and production experience in TypeScript, Go, Rust or a similar language.
• Experience building AI products, agents, model integrations or workflow orchestration systems.
• You write clean, tested and maintainable production code.
• A sound understanding of system design, APIs, infrastructure, performance and monitoring.
• Comfort working through ambiguous technical problems and owning the outcome end to end.
• Strong written and verbal communication.
• An interest in electronics, hardware design or PCB engineering, and the ability to pick up an unfamiliar technical domain quickly.

Previous PCB design experience is useful and not required. Nor is a PhD or a publication record: a degree in computer science, engineering or a related field is welcome, and evidence that you have built and trained these systems counts for just as much.

Nice to have

None of this is required. It is what would make the work easier from the first day, not what decides the first email.

• Publications or preprints at a major ML venue
• Reinforcement learning for combinatorial optimisation
• Graph neural networks, geometric deep learning or physics-informed models
• Structured generation, constrained decoding or grammar-based sampling
• Distributed training, quantisation or inference optimisation
• Experience building developer tools or open-source software
• Familiarity with KiCad or another EDA tool
• Contributions to technically ambitious open-source projects
• Prior experience at an early-stage startup