Please submit your CV in English and indicate your level of English proficiency.
Mindrift connects specialists with project-based AI opportunities for leading tech companies, focused on testing, evaluating, and improving AI systems. Participation is project-based, not permanent employment.
What this opportunity involves
We're building competitive-programming problems and their automatic graders (test cases + checkers) used to train advanced AI coding models. You'll create problems, write both correct and deliberately-wrong solutions, and design the tough test cases that catch subtly-broken code. Your work is what makes the training signal trustworthy.
To apply, share: your competitive-programming handle(s) + current/peak rating, any problem-setting experience, and IOI/ICPC/olympiad history.
What you'll do
Write original algorithmic problems — clear statement, sound constraints, an intended solution;
Write correct solutions in C++ and Python;
Write realistic wrong solutions (common mistakes) to test against;
Build and harden test cases: generators, edge cases, stress and "hacking" tests;
Write checkers/interactors (testlib) for problems with more than one valid answer;
Confirm the quality bar: correct solutions pass, wrong ones fail, within time limits.
What we look for
Strong competitive programming background, shown by a public profile — a Codeforces rating (see levels below), OR an IOI / ICPC / national-olympiad record;
Fluent contest C++ (STL, complexity); comfortable writing Python;
Can explain why a solution is wrong and build an input that breaks it;
Strong written English (C1+).
Levels (by Codeforces rating, or equivalent olympiad achievement):
Associate — 1700–2100 (Expert / Candidate Master): easier problems;
Expert — 2200–2600 (Master / Grandmaster): harder problems
Senior / Reviewer — 2600+ (Int'l Grandmaster) or IOI/ICPC medalist: hardest problems + quality review.
Nice to have
IOI / ICPC alumni and top competitive programmers especially encouraged to apply;
Experience setting or testing problems for real contests (Codeforces rounds, ICPC, national olympiads, online judges);
Testlib experience (checkers, validators, generators);
Experience creating data for LLM code benchmarks or similar RL / evaluation datasets.
How it works
Apply → Pass qualification(s) → Join a project → Complete tasks → Get paid
Project time expectations
For this project, tasks are estimated to require around 10-20 hours per week during active phases, based on project requirements. This is an estimate, not a guaranteed workload, and applies only while the project is active. Tasks must be submitted by the deadline and meet the listed acceptance criteria to be accepted.
Compensation
Paid per accepted task. Your rate depends on the qualification tier you reach and how efficiently you complete tasks — up to the equivalent of $90/hr. Because payment is per task, a faster pace raises your effective hourly rate.
Mindrift, powered by Toloka, is a platform that connects domain specialists and experts with AI project opportunities from major tech innovators. The company focuses on post-training and evaluation of frontier AI models by creating domain-specific reinforcement learning environments, tasks, and evaluation frameworks. Mindrift enables experts across fields—including mobile development, management consulting, physics research, and other domains—to shape how next-generation generative models learn and perform by converting real-world expertise into structured learning environments. The platform operates on a project-based model, allowing contributors to work on diverse AI initiatives including mobile app development, consulting domain training, physics problem design, and other specialized tasks for leading technology companies.
