"'AI-First' Begins with Recruitment... Krafton Assesses AI Competency with In-House Solution"
Development of AI-Native Recruitment Solution "Cofa-Probe"
Structuring and Evaluating the AI-Applicant Collaboration Process and Problem-Solving Competency
Addressing LLM-Based Assessment Limitations with "Meta Harness"
Krafton announced on August 31 that it has developed “Cofa-Probe,” an AI-native recruitment solution that structurally evaluates the collaborative process between artificial intelligence (AI) and job applicants. The company recently verified its effectiveness at a hackathon. Following its declaration last year of being an “AI-first” company, Krafton has established a system that assesses AI competency in the hiring process based not only on outcomes but also on the process and rationale behind them.
Participants are working on assignments at Krafton's AI native hackathon, 'Copaton,' held last month on the 30th at 'Pubgiseongsu' in Seongsu-dong, Seoul. Provided by Krafton
View original imageWhile traditional recruitment assessments have only examined completed results, Cofa-Probe observes the entire process in which applicants work with AI to define problems and validate outcomes. It virtually recreates environments similar to the actual work of AI engineers, evaluating how candidates interact with various work agents to identify and solve issues. Specifically, it analyzes and structures the full spectrum of AI collaboration — from defining problems, assigning roles and giving instructions to AI, searching for supporting evidence, revising decisions, recovering from errors, and validating results.
During the evaluation process, Cofa-Probe automatically collects collaboration records such as conversation logs, prompts, tool settings, and evidence of verification executions, along with the source code. After each task is completed, applicants participate in a post-task Q&A session to explain their deliverables and decision-making process. An interview reference report is then generated based on assessment criteria for each competency.
The system also addresses limitations of large language model (LLM)-based evaluations. When using LLMs, the same submission can yield different results each time it is executed, and differences in expression may obscure meaningful distinctions in skill. To resolve this, Krafton applied “Meta Harness” technology, which repeatedly tests and improves the execution structure of its AI evaluation agent. By leveraging OpenAI’s Codex, the company simultaneously achieved both reproducibility in assessing identical submissions and differentiation between distinct ones.
Krafton introduced Cofa-Probe at the “Kopaton” AI-native recruitment hackathon, held on July 30 at “PUBG Seongsu” in Seongsu-dong, Seoul, and completed its real-world validation. The event was divided into the “Krafton FDE Track” and the “CJ Olive Young AI Engineer Track.” Participants chose one of the practical tasks presented by either company, aiming to solve efficiency issues in the actual work environment using AI.
During the build process, Krafton used Cofa-Probe for real-time assessment and held an open discussion session with participants based on blind evaluation reports. As a result, Krafton reported that its evaluation principles were successfully applied to both its own task, which emphasizes problem definition and communication, and CJ Olive Young’s task, which reflects the context of real-world commerce operations.
Krafton sees significant meaning in implementing Cofa-Probe in the field, as it positions AI not merely as a tool to replace human judgment, but as a system to support better problem-solving and decision-making. Above all, the company succeeded in evidencing observable data of candidates’ AI capabilities, which are difficult to confirm through the final outcome alone.
Hot Picks Today
"Two Spend 500,000 Won in a Flash, Yet Wait Two Hours?" ... 'Dreamland' Shunned by Kids and Foreign Tourists
- The Era of Getting 200 Million Won After 30 Years Is Over... A Japanese Bank's "Perpetual Department Head" Who Just Clocked In Cries Out
- Spain Hit by Its Worst Tragedy... Lee Kangin, Moreno and Real Madrid Join Soccer World in Mourning
- They Said, “Make It Your Home for Life” - Then Suddenly Kicked 200 Passengers Off... What Happened to the Around-the-World Cruise?
- "130 Rice Bowls Sold Out in a Day"... 'Unfiltered Face' Adds 8 Million Won to Night Market Sales: China's Bungeunwoo
Based on operational experience and on-site feedback gained at Kopaton, Krafton plans to further refine the assessment criteria and agent environment of Cofa-Probe. Jaemin Park, Head of AI Frontier at Krafton, stated, “The competitive edge in the AI era comes from defining problems with AI, verifying evidence, and rigorously validating results to the end. We will continue to enhance Cofa-Probe and establish a new standard for assessing AI-native talent.”
© The Asia Business Daily. All rights reserved. Unauthorized AI training and use prohibited.