Open role

AI Training Engineer — Coding Agent Evaluation.

Contract, remote · 10–40 hrs/week · paid hourly in USD, rate agreed before any assessed work · a new bench being built now, no guaranteed start.

The role

What the work is.

You will not be writing production code. You will be judging whether a model reasons like a good engineer: reviewing agent trajectories, writing reference solutions, authoring repository-level tasks, and building graders that do not fall over when a model games them.

Requirements

The bar.

Eight or more years. A senior title held. Direct full-time engineering history. Clear written English. Comfortable with current AI coding tools.

Stated plainly

Where the pipeline actually stands.

This is a new bench, not an existing one with an opening waiting to be filled. We are lining up client engagements now, and there is no guaranteed start date — if you pass the screen, you may be the first person on it rather than the next in an established queue. Engineers in this market are routinely screened and then left waiting with no explanation. We would rather tell you the truth: if there is no work the week you finish the screen, we will say so directly, and tell you when that changes.

Commitment

A human replies to everyone.

Every applicant gets a reply from a person, not an automated rejection. That includes a no. It includes “there is no work right now” if the pipeline is empty when you apply.

Commitment

Paid hourly, in USD.

The rate is agreed with you directly before you do any assessed work, in USD, paid hourly — never announced after the fact, and never a range you discover you're at the bottom of. Approved invoices are paid within 14 days. Not “when the client pays.”

Next step

How to apply.

The application below covers your background, availability, and rate expectations, plus your resume. A person reads every submission and replies — including if the honest answer is that there is no work for you right now.

PDF or Word document, up to 3MB.