AI training jobs for software engineers
Coding models are only as good as the people who check their work. If you write software for a living, several AI training platforms will pay you to review, test and break AI-written code. Here's who's hiring, what each listing pays, and what they ask for. Every listing below was checked on November 2, 2026.
Last checked: November 2, 2026
The short answer
- Coding roles are open right now at DataAnnotation, Mercor, Alignerr, Meridial, Mindrift and Turing. Outlier also recruits coding experts but doesn't publish rates. [source][source]
- Posted pay on November 2, 2026 ran from $8 to $150+ an hour (USD), depending on the listing. These are listed rates and ceilings, not typical earnings. Most of the higher figures are ceilings ("up to").
- The work is judging code, not writing tutorials. You run coding sessions with an AI model, check whether its code actually works, compare answers, and explain what went wrong.
- Expect a test first. DataAnnotation's starter assessment can be taken only once, Mercor uses a short AI interview, and Mindrift asks for a technical assessment. [source][source][source]
Coding AI training roles open now
| Platform | Role | Posted pay (USD) | What they ask for |
|---|---|---|---|
| DataAnnotation | Software Engineer (coding projects) | "Up to $150+/hr" (DataAnnotation's wording); its FAQ says coding projects start at $40–$150+ per hour | Professional or serious open-source experience shipping production code; most tasks use Python, C++, Rust, Go or JavaScript; no degree required [source] |
| Mercor | Software Engineering Expert (talent network) | Listings have posted $70–$150/hour, set per project | Professional experience building and shipping software; clear written communication [source] |
| Alignerr | Software Developer: build your own project with an AI coding assistant | $60–$100/hour | A software project you personally own, in a git repository; comfortable with the terminal and git [source] |
| Meridial | Coding Specialist – AI Trainer (US) | $8–$65/hour; exact rate set by experience, expertise and location | Computer science and software engineering depth; open-source work, technical writing or competition results help [source] |
| Mindrift | Coding (code review, generation, automation) | "Up to $5,400/month" (Mindrift's home page) | A technical assessment of your coding skills [source] |
| Turing | LLM Trainer – Software Engineering (SWE Bench) | Not shown on the listing | Bachelor's or master's in computer science or a related field; 3+ years of professional software engineering; at least 4 hours a day and 20 hours a week, with 4 hours' overlap with Pacific time [source] |
Pay is what each listing showed on November 2, 2026. Listings open, close and change rate often, so check the listing before you count on a number. "Up to" figures are ceilings, not what most people earn. Mindrift's figure is monthly, not hourly.
What the work actually looks like
- Running agentic coding sessions. DataAnnotation's Software Engineer role has you run AI coding sessions on real engineering tasks, analyze what the model produces, and red-team it to expose unsafe code, faked test passes and confidently wrong solutions. [source]
- Grading code and design. Mercor's software engineering pool grades whether model-written code handles edge cases and whether its tests really exercise what they claim, and flags code that would pass review but fail in production. [source]
- Building on your own project. Alignerr's listing pays developers to work on a project they own with an AI coding assistant, choosing the better of two responses at each step and explaining why. [source]
- Writing benchmark tasks. Turing's SWE Bench role designs realistic engineering tasks for frontier models, with test cases and solution explanations. [source]
- Quizzing the model. Meridial's coding specialists talk the model through coding and computer-science questions, check its answers, and record reproducible error traces. [source]
How to qualify
- Shipping experience matters most. DataAnnotation and Mercor both ask for professional experience shipping software; DataAnnotation says no degree is required. Turing asks for a degree and 3+ years. [source][source][source]
- Know a mainstream language well. DataAnnotation says most tasks use Python, C++, Rust, Go or JavaScript. [source]
- Treat the first test as final. DataAnnotation's starter assessment has no retakes, so read the instructions and review before you submit. [source] More tips: How to get accepted on AI training platforms.
- Write clearly. DataAnnotation calls your written explanations "the training signal," and Alignerr asks for short, honest evaluations of each choice. [source][source]
Tips
- Apply to more than one. Projects start and pause with client demand, so one platform may not have work for you right away.
- Check the time commitment. Turing's listing asks for at least 4 hours a day and 20 hours a week with Pacific-time overlap; Alignerr's and DataAnnotation's are flexible. [source][source][source]
- Read the project rules. Alignerr's own-project role excludes work owned by an employer, client or university, projects under NDA, and projects that process other people's personal data, and says session transcripts can't be withdrawn once submitted. [source]
- Know when you get paid. Pay days differ a lot by platform: see When do AI training platforms pay?
FAQ
Do I need machine learning experience?
No. The listings above ask for software engineering skill, not ML. Alignerr's own-project role actually excludes ML infrastructure and model-training projects as the project you bring. [source]
Which pays the most?
On posted figures, DataAnnotation ("up to $150+/hr") and Mercor ($70–$150/hour on its listings) are at the top. Treat both as ranges and ceilings: your rate depends on the project and your test results. [source][source]
Can I do this outside the US?
Some of it. Alignerr runs the same own-project listing for developers in cities from Tokyo and Seoul to Kraków and Mexico City. Meridial's coding role is US-only. Check our countries page for each platform. [source][source]
Is this the same as the cybersecurity roles?
No. Security roles focus on exploits and red-teaming. See AI training jobs for cybersecurity professionals.
Try it
See Mercor (affiliate link: we may earn a fee if you sign up, at no cost to you). Or compare coding pay across platforms on our pay page.
Pay and listings checked November 2, 2026. Rates and openings change often; check each listing before you apply.