Backend Coding Task Creator and AI Evaluator

The work

You will create challenging backend coding tasks for AI systems to solve and assess the solutions they produce. The work combines backend engineering judgment with structured review of code quality.

You will also build deterministic verifiers. These checks should accept reasonable, valid solutions and reject code that is incorrect or incomplete.

• Create realistic backend feature-development and software bug-fix tasks.
• Build verifiers that test whether solutions are correct and complete.
• Review code for correctness, performance, automated testing, and maintainability.
• Identify debugging issues, software bugs, system bottlenecks, and performance problems.
• Use engineering judgment to create challenging, high-quality AI training examples.

What it pays and takes

This is a part-time contractor role with less than 20 hours per week. The work is open to candidates in the countries listed for the project and requires English fluency.

• $30 to $100 per hour equivalent, paid per task that meets the project specifications.
• Part-time work of less than 20 hours per week.
• Open to candidates across the Americas, Europe, Asia, Africa, and the Middle East, including the countries listed in the application.
• Professional backend development experience with Java, Node.js, Go, Rust, TypeScript, Python, C++, or a similar language.
• Experience building scalable services and APIs.
• Strong knowledge of algorithms, data structures, and foundational computer science.
• Ability to refactor code and implement complex features in large-scale applications.
• Strong analytical, problem-solving, and technical communication skills.
• Experience with service testing, automated test suites, flaky-test diagnosis, testing practices, or CI test pipelines is helpful.

How it works

Apply on OpenTrain with your resume, then complete the application on the hiring site.

About AI training work

AI training is the human work used to improve artificial intelligence, including creating coding examples and reviewing model responses. Experienced software engineers are needed to judge whether generated code is correct, reliable, and useful.