What you would do

Write bug-fix and feature tasks in real open-source codebases, such as a race condition in a background worker, with hidden tests that decide pass or fail.
Review model-written pull requests line by line for correctness, security and edge cases, and explain each issue so another engineer can check it.
Build terminal tasks, such as a broken Docker build or a failing CI step, with a script that verifies the fix.
Write algorithm problems with test cases that catch slow or subtly wrong solutions.
Compare two model answers to the same coding question and say which is better, and why, in a few precise sentences.
Write reference solutions and grading rubrics that two reviewers would apply the same way.

What we look for

You study at or graduated from a college or university. Internships are not required.
One substantial coding project you built, on your own or with AI tools, that you can explain in depth: what it does, how it works and what you would change.
You can read unfamiliar code in a common language, such as Python, TypeScript, Java or C++, and explain exactly why something fails.
You test your own work: you reproduce bugs, write tests and check edge cases before calling something done.

A degree in computer science or a related field; a master’s or PhD is even better.
Professional engineering work, especially at an AI lab or AI company.
Code other people rely on: a shipped product, an open-source project or a library you maintain.
Depth in more than one area, such as backend and security, or systems and performance.
Experience reviewing other people’s code, as a maintainer, team lead or teaching assistant.

Task size: Tasks take about 30 minutes to 5 hours. Each one shows its estimated time and price before you start, and you are paid that price even if it takes you more or less time.
Pay: Per task accepted in review: its estimated time times the project’s rate. A task estimated at 2 hours at $120 an hour pays $240. Accepted tasks are paid within 30 days.
Hours: Projects last 1 month to a year. Each sets a weekly minimum, usually about 20 hours of estimated work, and you choose when you work.
Tools and review: Your own computer with a terminal, git and the languages each task names. Every task is reviewed against the project’s written quality standard before it is accepted.