An older role: first posted 36d ago, and AI expert network still listed it when we checked today. Newer roles tend to fill faster. See the jobs hiring now.
What the work is
About the role
A frontier engineering-reasoning evaluation run in collaboration with a leading AI research lab. The work measures whether state-of-the-art models can reason from first principles in your engineering domain rather than retrieve facts from training data — and you get visibility into the model's internal reasoning on your own tasks.
What you'll do
Domain experts write simulations and specification sheets; the model attempts to design an artifact — controller gains, circuit parameters, geometry — that satisfies every spec.
You define the simulation, a set of specs with pass/fail thresholds, and the prompt. The model probes your simulation with a limited number of calls, then submits a final design. An agentic grader runs the simulation and scores spec satisfaction.
Who gets hired
Who we're looking for
- Deep applied expertise in control systems, analog or RF circuits, power electronics, or mechanical design — PhD or equivalent industry track record.
- Comfortable writing simulations in Python and expressing acceptance criteria as quantitative, testable thresholds.
- Able to design a problem you can't shortcut yourself, but can still judge rigorously.
- Comfortable working in a browser-based studio and GitHub.
Support
Onboarding calls run daily, alongside internal tooling built to help you work faster. Prior model-evaluation experience is not required.
Pay
$60–90/hr, fully remote.