Paid, remote contract work turning published robotics and reinforcement learning research into coding tasks for frontier AI agents. For PhD-level researchers and research engineers who publish and ship code.
ML Engineer work for AI training
This is a talent pool, not an open role. Applying adds you to a roster this employer draws on when work comes up. There may be no specific vacancy to fill right now, and being accepted does not guarantee an assignment.
Remote (Global)
$20 – $100/hr
October 30, 2026
Job description
Design the evaluations and rubrics that show where models go wrong. Build evals, write rubrics, and review training data and model pipelines.
What you would do
- Design evaluations and benchmarks.
- Write rubrics that turn judgment into clear criteria.
- Review training data and model pipelines for errors.
Who we look for
- Hands-on ML or data science experience.
- A degree in ML, AI, data science or a related field, or equivalent work.
How It Works
Apply once, work in your field.
Remote, on your own hours.
01
Apply
Tell us your field and background. It takes about a minute.
02
Assessment
Take the short assessment for your field.
03
Matching
When a project in your field opens, we match people by field and results.
04
Paid work
You are paid for each accepted task at the project's rate.
Next step
Ready to apply?
Pick ML Engineer as your field. You can add other fields you know well on the same form.
Apply now
Opens experts.rustlabs.ai in a new tab. If they ask how you found the role, mention RemoWork.
Paid, remote contract work turning published ML systems and GPU kernels research into coding tasks for frontier AI agents. For PhD-level researchers and research engineers who publish and ship code.
Gramian Consultancy is a boutique consultancy specializing in IT professional services and engineering talent solutions. With a strong background in software engineering and leadership, we help companies build high-performing teams by matching them with professionals who truly…
We are building LLM evaluation and training datasets to train LLM to work on realistic software engineering problems. One of our approaches in this project is to build verifiable SWE tasks based on public repository histories in a synthetic approach with human-in-the-loop; while…