In September 2026, hh.ru launched its first School of Analysts to train students for future internships. The company received over 250 test assignments for 25 available spots. The selection process focused on practical data interpretation skills rather than specific programming languages, with over 85% of applicants choosing Python.
To identify candidates using AI tools without proper understanding, the organizers included specific traps in the test tasks. These included 'penalty' traps for data errors and 'signal' traps designed to catch superficial AI-generated responses. Only 6% of candidates received full marks for the recommendation section of the assignment.
The selection process involved two main tasks: a statistical analysis of a synthetic dataset containing 100,000 job postings and a critical review of a social media post about job hunting. Only 13% of candidates successfully completed the data cleaning and methodology description for the first task.
The organizers utilized AI models for automated grading to optimize resources, as manual review of all 250+ assignments would require over 100 man-hours. The final selection relied on machine-adapted markings combined with selective human oversight to classify candidates.
