The work
You will review code patches generated by large language models for open source repositories. Your assessments will help improve AI systems by showing whether proposed changes are correct, useful, and suitable for the repository.
You will also review submissions from other subject-matter experts. Clear written reasoning is a central part of the work.
- Review AI-generated code patches for repositories you maintain.
- Judge whether proposed changes are correct, useful, and appropriate.
- Write detailed feedback that clearly explains each assessment.
- Review work submitted by other software engineering experts.
- Evaluate changes across programming environments such as Python, C, C++, Java, TypeScript, Rust, and Go.
What it pays and takes
This is a globally remote contractor opportunity with flexible scheduling. The expected availability is 15 hours per week, and payment is based on accepted tasks that meet project specifications.
- Pay: $150 to $300 per hour equivalent, paid per task that meets project specifications.
- Availability: About 15 hours per week.
- Work arrangement: Remote contractor work with flexible scheduling.
- Language: Fluent in English.
- Location: Open to candidates in AE, AR, AT, BD, BE, BH, BR, CA, CH, CL, CO, DE, DK, DZ, EG, ES, FR, GB, GH, GR, ID, IN, IT, JO, JP, KE, KG, KW, KZ, LB, LK, MA, MX, MY, NG, NL, OM, PE, PH, PK, PS, PT, QA, SA, SG, TH, TN, TW, US, UZ, VN, and ZA.
- Experience: The listing is marked entry level, while the work requires senior-level software engineering judgment.
- Requirements: Active or demonstrable open source repository maintenance experience, a clear GitHub or GitLab profile, experience reviewing contributor code, and strong technical writing skills.
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 behind systems that generate and understand content. Contributors review examples, assess model responses, and provide feedback so AI systems learn to produce more useful results. Software engineering expertise is valuable when models need careful review of technical code and decisions.