AI coding assistants learn from examples of good code, bad code and the reasoning between them. Companies pay developers to write those examples, review model output and build hard problems the model can't solve yet. It's one of the better-paid corners of AI training work.
Here's what the job involves, where to apply, how screening works and what to watch out for.
What AI coding trainers do
- Write coding problems with tested solutions. Realistic tasks, from a single function to a multi-file bug fix, with tests that prove the answer works.
- Review AI-written code. Find the bug, the security hole, the edge case it missed, and explain it.
- Compare two solutions and say which is better and why: correctness first, then clarity, performance and style.
- Grade reasoning. Check whether the model's step-by-step explanation actually matches what the code does.
- Work in real environments. Some projects use terminals, repositories and test suites rather than chat boxes.
Most of your output is judged on two things: does it run correctly, and is your written explanation clear.
Who it suits
- Working software engineers in common languages such as Python, JavaScript or TypeScript, Java, C++, Go or Rust.
- Data scientists and ML engineers, especially for notebook, SQL and data-pipeline tasks.
- Specialists in areas that are hard for models: systems programming, security, DevOps, game development, scientific computing.
- Strong computer science students and recent graduates. Some projects accept them, especially for review work.
You don't need AI experience. You need to read code carefully and explain problems precisely.
Where to apply
- Turing is strongest for coding. It hires software developers, data scientists and engineers, including LLM trainer roles. Its process: build a profile, get matched, then take skill assessments and live interviews. Its jobs page asks for strong written and spoken English.
- Mercor runs coding roles alongside its other expert work.
- micro1 lists engineering among its fields and screens with an AI interview.
- Handshake AI has software engineering roles for its fellows. Most of its projects must be done from inside the US.
For a side-by-side view of the three biggest, read Mercor vs micro1 vs Turing. For this week's hiring, check the tracker.
What it pays
Pay is set per listing and depends on the role and, on some platforms, on your country. We only quote figures from a platform's own page. Handshake AI's page shows roles from $40 an hour up to $125 an hour, and lists a game developer role at the top of that range.
Turing has been reported to pay trainer roles at lower rates in some countries than its headline engineering figures. Those reports come from review sites, not Turing itself, so check the rate on the actual listing.
How screening works
Expect more than a chat. Coding screening usually combines:
- An AI or live interview about your experience. Be ready to discuss real projects in depth.
- A coding test or take-home task, sometimes timed.
- Screen sharing. Mercor's docs say some AI interviews require full-screen sharing and some include a whiteboard.
Tips:
- Treat the test like code review at work. Read the whole prompt. Handle edge cases. Add tests if allowed.
- Explain as you go. Your reasoning is scored, not only the final answer.
- Use the language you're strongest in when you get a choice.
- Don't paste in AI-generated solutions. Platforms look for it during screening and on projects. Mercor's conduct policy names misuse of AI tools as something it reviews.
The catches
- Hours aren't guaranteed. Being accepted puts you in a pool. Projects end, sometimes suddenly. If that's you, read passed the interview but no projects.
- Some projects want long, fixed hours. Our research found reports of Turing engagements asking for full days with several hours of overlap with US Pacific time. That's a real job schedule, not a side gig. Check before you accept.
- Monitoring. Some platforms use time-tracking software that takes screenshots while you work, and ban VPNs.
- Confidentiality. Project code and prompts are confidential. Don't post them on GitHub or social media.
- Your day job. Check your employment contract for outside-work or IP clauses before you start.
Getting set up
You'll need a reliable computer that can run the tools a project uses, a stable connection and a quiet place for the interview. See what you need to work from home in AI training.
Where to go next
Find a coding role that's hiring now on the job list, and check the tracker for which platforms are adding roles this week.
Some Apply links are referral links: we get a bonus if you're hired, it costs you nothing and the application is the same.