AI Trainer (SWE)

What You'll Do6

• Design software engineering tasks that test the limits of advanced AI models.
• Write task instructions, set up reproducible environments, and craft reference solutions, tests, and evaluation criteria.
• Review model-generated code and technical outputs for correctness, robustness, and completeness.
• Spot problems in implementation, debugging, dependencies, performance, testing, or reproducibility.
• Establish objective standards that define what counts as a correct solution.
• Evaluate long-horizon or agentic coding submissions when the project calls for it.

Requirements5

• Paid experience training, evaluating, benchmarking, or improving AI systems.
• Strong hands-on Python skills.
• Professional experience with at least one more language such as Java, JavaScript, TypeScript, Rust, C++, C#, or Go.
• Solid grasp of software design, debugging, testing, version control, and dependency management.
• Public evidence of skill: substantial GitHub or GitLab repos, open-source contributions, maintainer roles, competition results (Codeforces, ICPC, Olympiad), developer tools, or technical write-ups.

Preferred Qualifications6

• Experience creating AI-training tasks, reference solutions, tests, graders, or benchmark problems.
• Experience reviewing other AI trainers' technical submissions.
• Codeforces rating around 1900 or higher, or strong results in another programming competition.
• Familiarity with Linux, Docker, CI/CD, command-line tools, or sandboxed environments.
• Prior work with long-horizon or agentic coding evaluations.
• Available to start immediately.

Who Should Apply

Software engineers who have already done paid AI-training or model-evaluation work and can show public proof of their coding skill. You should be comfortable with Python and at least one other language, and ready to commit 15+ hours per week on a flexible, remote schedule.

Salary Insight