Competitive Programming AI Training Expert

The Work

You will solve self-contained coding and technical reasoning challenges used to evaluate advanced AI models. The work involves algorithmic problem solving, debugging, constrained implementation, and reproducible software development.

• Solve coding, logic, debugging, and technical reasoning challenges with clear objectives.
• Explore unfamiliar codebases or terminal environments and decide how to approach each problem.
• Turn requirements into working solutions using suitable algorithms and data structures.
• Test solutions against constraints, edge cases, complexity expectations, and correctness requirements.
• Produce reproducible final results that can be evaluated consistently.

What It Pays and Takes

English fluency is required.

• $60 to $120 per hour equivalent, paid per task that meets project specifications.
• Approximately 15 hours per week, with a flexible schedule.
• Fully remote, with work available to candidates in the listed countries across the Americas, Europe, Asia, Africa, and the Middle East.
• Fluent in English.
• Strong understanding of common data structures and algorithms.
• Careful reasoning about constraints, complexity, edge cases, and correctness.
• Ability to debug unfamiliar code and resolve technically ambiguous problems.
• A public or otherwise verifiable Codeforces profile showing competitive programming experience.
• Practical coding ability in Python or C++.
• Competitive programming, competitive C++ events, judging experience, or recognized programming competition achievements are helpful.

How It Works

Choose the hours and days you work, including weekends if desired. To apply, submit your resume on OpenTrain and then complete the application on the hiring site.

About AI Training Work

AI training is the human work behind modern AI systems, including writing, testing, and reviewing examples that help models produce better results. People with strong technical skills are paid to solve difficult problems and judge whether model outputs are correct, useful, and reproducible.