For engineers

Engineering is one of the largest surface areas AI is being deployed against, which means it is one of the largest surface areas AI needs to be trained against. Software, systems, mechanical, electrical, civil, chemical, aerospace, and industrial engineering all have the same shape: a body of well-established practice, a set of hard-won intuitions about what works and what fails, and constant edge cases the textbooks do not cover.

Models are learning these fields, but they are learning them from documents. Documents are not enough. What they need next is working engineers correcting their reasoning on the exact problems working engineers see every day.

What the work looks like

• Response evaluation. You review model outputs on engineering prompts (architecture questions, debugging traces, design reviews, failure analysis) and score them against what a working engineer would say.
• Problem authoring. You write realistic engineering scenarios at the level of a real project, with enough detail that a model can learn from your reasoning about how you would attack them.
• Adversarial testing. You construct problems where the model is likely to give a confident, textbook-correct answer that would fail in a real production or field setting.
• Rubric review. You audit the criteria the lab is using to grade engineering outputs and flag anything that would let unsafe or non-standard practice through.

The work is asynchronous. Sessions typically run in blocks of two to four hours. Deliverables are structured.

What labs actually look for

Recent, active work in a specific engineering discipline. Formal training helps. What matters more is verifiable current practice and the ability to explain your reasoning cleanly in writing.

They do not look for AI experience or machine learning background. They look for depth in the discipline itself.

Where the pay lands

Rates are in USD. For engineers in emerging markets, this is often significantly above local compensation. For engineers in higher-cost markets, it is competitive with consulting or advisory work on an hourly basis.

How to enter

Sign up, upload a recent CV that makes your discipline and current work visible, and let the platform surface open roles in engineering evaluation, adversarial testing, and problem authoring. Apply where you can deliver on the depth. Skip where you cannot.