Competitive Programming Expert AI Trainer📣 Job Ad

About Mindrift and the Role

Mindrift connects specialists with project-based AI opportunities for leading technology companies, focusing on the testing, evaluation, and improvement of AI systems. This is a freelance, project-based role as a Competitive Programming Expert - AI Trainer, and it does not constitute permanent employment. The primary objective of this position is to develop competitive programming problems and their corresponding automatic graders, including test cases and checkers, which are essential for training advanced AI coding models. Your contributions will directly ensure the reliability and quality of the training data.

Key Responsibilities

• Write original algorithmic problems, including a clear statement, sound constraints, and an intended solution.
• Develop correct solutions in C++ and Python.
• Create realistic wrong solutions that simulate common programming mistakes for testing purposes.
• Build and refine test cases, incorporating generators, edge cases, stress tests, and "hacking" tests.
• Develop checkers and interactors using testlib for problems with multiple valid answers.
• Ensure quality standards are met, verifying that correct solutions pass and incorrect solutions fail within specified time limits.

Required Qualifications

• A strong competitive programming background, evidenced by a public profile (*, a Codeforces rating) or an IOI / ICPC / national-olympiad record.
• Fluency in contest C++ (STL, complexity) and comfort in writing Python.
• Demonstrated ability to identify why a solution is incorrect and construct an input that exposes its flaws.
• Strong written English proficiency (C1+).

Desired Skills and Experience

• Prior participation as an IOI / ICPC alumni or recognition as a top competitive programmer.
• Experience in setting or testing problems for competitive programming contests (*, Codeforces rounds, ICPC, national olympiads, online judges).
• Familiarity with testlib for developing checkers, validators, and generators.
• Experience in creating data for Large Language Model (LLM) code benchmarks or similar Reinforcement Learning (RL) / evaluation datasets.

Project Structure and Compensation

This role operates on a project basis. Upon successful application and qualification, you will join a project to complete assigned tasks. Tasks are estimated to require approximately 10-20 hours per week during active project phases; however, this is an estimate and not a guaranteed workload. All tasks must be submitted by their deadlines and meet specified acceptance criteria to be approved. Compensation is provided per accepted task, with rates varying based on your qualification tier and task completion efficiency, up to $90 per hour.

Qualification Tiers

Candidates are evaluated and assigned to qualification tiers based on their Codeforces rating or equivalent olympiad achievements:

• Associate: Codeforces rating of 1700–2100 (Expert / Candidate Master), focusing on easier problems.
• Expert: Codeforces rating of 2200–2600 (Master / Grandmaster), focusing on harder problems.
• Senior / Reviewer: Codeforces rating of 2600+ (International Grandmaster) or an IOI/ICPC medalist, focusing on the hardest problems and quality review.