Put your open-source maintainer judgment to work, remotely. You design realistic engineering tasks from the codebase you know best, write reference solutions, and review AI-written code, tests and rubrics against your project's real quality bar. Mostly asynchronous, so evenings and weekends work, but it needs 30-40 hours a week.
What you will do
- Design realistic engineering tasks grounded in your open-source codebase, such as the features, refactors and bug fixes a maintainer would assign.
- Write accurate, well-structured reference solutions.
- Review AI-written code, tests and grading rubrics against the repository's architecture, conventions and quality bar, with clear written feedback.
- Build, run and test the project locally, including with Docker, to confirm tasks, solutions and tests behave as intended.
- Write precise rationales on what a strong solution looks like and where AI-generated changes fall short.
- Help research and engineering teams close knowledge gaps in software engineering, code review and open-source maintenance, and work with other experts to keep the work consistent.
What you need
- Maintainer status (named in MAINTAINERS or CODEOWNERS, or 10+ PR reviews in the last 12 months) or core contributor status (3+ substantive merged PRs in the last 12 months) on a widely used open-source project.
- That repository is 2+ years old, has real users, an active code review culture, and is mostly human-written.
- Years of sustained open-source contributions, verifiable on your own GitHub account.
- Senior-level code review judgment, including critical review of AI-written code, tests and edge cases.
- Hands-on ability to build and run the project and its test suite locally, including with Docker.
- Strong written communication and the ability to explain complex technical decisions clearly. You write all rationales and feedback yourself, in your own words.
- Authorization to work in the country you live in.
Good to know
- The role needs 30-40 hours a week as your main work, so it does not fit next to a full-time job.
- Most work is asynchronous, with occasional meetings with the lab's researchers.
- Exclusivity: during the engagement, no work for another AI lab or similar AI data company, including other projects on the same platform.
- Before starting you complete background checks and confidentiality agreements, and you work on a company-provided device.
- You are engaged through a staffing company that acts as employer of record, with the chance to be placed in an AI lab's extended workforce.
How pay works
This job is paid hourly, at the rate listed above, for 30-40 hours a week.
Apply (opens in a new tab)