AI Training Jobs for Software Engineers: 374 Roles Paying $40 to $150 an Hour

As of 4 November 2026, AI training jobs for software engineers number 374 open roles across 27 AI training platforms, with typical pay running from $40 to $150 an hour and a median of $65 an hour, according to a live count checked against each platform's own job board every hour. AI training jobs for software engineers are hourly contract roles in which engineers review and rank the code a model writes, write coding tasks with the tests a correct answer must pass, and fix a model's failed attempts in real repositories. The highest advertised rate in the count is $400 an hour, at micro1, but most roles sit far below that ceiling, and 52 of the 374 roles state no hourly rate at all, so the headline figures describe only the roles that publish pay.

Count date	4 November 2026
Open roles for software engineers	374, across 27 AI training platforms
Typical hourly pay	$40 to $150 an hour, median $65
Highest advertised rate	$400 an hour, at micro1
Roles stating pay	322 of 374; 52 state no hourly rate
Rates measured	258 distinct advertised rates, across 22 platforms
Work type	Remote, hourly contract work; hours depend on the project
Usual first step	A coding assessment or a paid trial task
Application fee	None; a legitimate platform never charges you to apply

AI training jobs: what the work actually involves

The work splits into distinct kinds of tasks, and the count groups the 374 roles accordingly. Code review, evaluation and agentic coding account for 77 roles: reviewing, ranking and grading code a model writes, auditing the step-by-step traces of coding agents, and building tasks in the style of SWE-Bench from real repositories, with tests a fix has to pass. A further 95 roles are open to software engineers in any language, covering writing and fixing code to the tasks a project sets, checking a model's solutions and explaining what it got wrong. Python accounts for 33 roles, JavaScript, TypeScript and full-stack work for 48 roles, and other languages including Java, C++, Go, Rust and mobile development for 79 roles. A further 42 roles cover data science and machine learning, where engineers check a model's analyses, experiments and code and design the evaluations it is tested on.

Most roles start with a coding assessment or a paid trial task, and many ask for a GitHub profile or years of professional experience in the language they name. The count's own guidance adds a warning for employed engineers: check the outside-work and intellectual property clauses of an existing contract before starting, and never use an employer's code or credentials in a task. The work is usually part time and flexible, but hours depend on the project, and a project can end without warning.

AI training jobs: who pays what, platform by platform

The platform table shows how wide the pay spread is. SuperAnnotate lists 49 roles at up to $120 an hour, Alignerr lists 46 roles at $20 to $400 an hour, and Invisible lists 44 roles at $6 to $150 an hour. Micro1 lists 32 roles at $20 to $400 an hour, Anyone AI lists 32 roles at $15 to $150 an hour, and Turing lists 29 roles at $100 to $150 an hour. AfterQuery lists 26 roles at $45 to $175 an hour, Weekday lists 16 roles at $20 to $230 an hour, and Mercor lists 15 roles at $50 to $180 an hour. At the smaller end, RemoExperts lists 6 roles at $50 to $150 an hour, Surge AI lists 1 role at $100 to $150 an hour, and DataAnnotation lists 9 roles at $40 to $150 an hour.

Read those ranges with care. A range such as $20 to $400 an hour does not mean a typical engineer earns near the top of it; the count's own typical band of $40 to $150 an hour, with a $65 median, is the better guide, measured from 258 distinct advertised rates and excluding 64 roles that advertise only a maximum rate such as "up to $X an hour". The platform list is also ordered by what applying is worth, with a referral boost for platforms that pay the source a referral fee, so position on the list is not a neutral ranking of quality. Pay is set by the platform, not by the listing site.

The wider research behind the count puts software engineering in context. In the source's field-level research, software and IT counted 406 open roles on 3 November 2026 across 28 platforms, up 4.9 per cent, against 2,442 people with profiles in the field, or 6.0 people per role. That is a competitive market: many engineers are chasing each role, and the assessment stage is where most applications end.

AI training jobs: how to apply, and what to do next

Start on the platform's own job board, and confirm three things before spending time on an assessment: the language and stack the project actually wants, the country and credential rules for that project, and the pay terms for the specific role, including whether the rate is fixed or a range and how hours are approved. A strong engineer in the wrong stack or the wrong country will not pass the first screen. Keep a record of each application, the role text and the rate stated, because listings change and roles close without notice.

For engineers already in full-time work, settle the contract question first, since a conflict over outside work or intellectual property can cost more than the extra income is worth.

Finally, a legitimate platform never charges you to apply or to get work, and any message asking for a fee for training, equipment or onboarding is a warning sign. Growentix reported on 4 November 2026 that 1,100 AI startups were advertising 13,514 roles, and Indian AI hiring rose in October 2026 even as overall hiring stayed uneven. AI training work is the contract, hourly side of that same market: flexible and genuinely paid, but competitive and worth checking role by role on the platform's own page.

Sources: AITraining.jobs, "AI Training Jobs for Software Engineers", figures as of 4 November 2026, checked against each platform's job board; field-level research "Expert Fields in Demand for AI Training" and "AI Training Talent: Supply and Demand", data from 7 October 2026 to 3 November 2026.