Researcher, engineering capability

About Linewise

Linewise captures how physical things are made, from the first design to the final assembly, and turns it into training data for frontier AI and robotics models. Our long-term aim is Physical RSI: recursive self-improvement for the physical world, where AI helps engineers design and plan, robots help build and inspect, and each generation makes better machines, lines and factories for the next. This role is part of Linewise Research, which develops our methods for capture, quality control, annotation and model evaluation, and our agent-native data infrastructure. All of it is built around real industrial tasks.

About the role

You'll research how well AI models do real engineering work, from CAD and CNC programming to simulation and design for manufacturing, and how to make them better. You'll build evaluations from the engineering data we capture, find where frontier models fall short, and run post-training experiments to show which data moves the score. Your results decide which datasets we build and propose to frontier labs.

What you'll do

• Build evaluations for engineering tasks from our data: CAD modelling, CNC and CAM programming, simulation and FEA, and design review.
• Evaluate frontier and open models on them, and find out where and why they fail.
• Run post-training experiments on our data, and measure what improves on real tasks.
• Propose the datasets that would raise a model's score, and work with our engineers to capture them.
• Help set how we measure data and model quality across Linewise Research.
• Publish what you find as papers and benchmarks.

What we're looking for

• Research experience in model evaluation or post-training, such as fine-tuning or RL, shown in papers, open-source work or models you've shipped.
• Strong Python and PyTorch, and hands-on experience training or fine-tuning large models.
• Experience designing evaluations that hold up: clear tasks, reliable grading and results you can trust.
• An interest in engineering work, and in what it takes for a model to do it well.
• Clear written English. You can explain a result in a short write-up.

How we work

• You own your area end to end, from the first spec to the final delivery.
• You work closely with engineers, operators and researchers.
• English is the working language. We're based in San Francisco, with global operations.

How to apply

Send us two things:

• A short note on what you've built and what you'd like to own at Linewise.
• Your CV, or a link to your work.

Email them to info@linewise.io with the role title as the subject. The Apply button sets this up for you.