EBIT, an AI training data specialist led by CEO Yonghan Park, announced on January 23 that it has been selected as an official data supplier for the 2026 Data Voucher Support Program, a government initiative overseen by the Ministry of Science and ICT and operated by the Korea Data Agency (K-DATA).

The Data Voucher Support Program is designed to help small and medium-sized enterprises, startups, and research institutions develop data-driven services and products by subsidizing the cost of data acquisition and processing. Through this program, EBIT plans to provide advanced AI training datasets focused on mathematics and high-level coding to support the development of next-generation large language models (LLMs) and domain-specific AI systems.

EBIT’s datasets consist of high-difficulty training data covering mathematical reasoning, algorithmic thinking, and problem-solving-oriented code generation and analysis. Rather than simple question-and-answer formats, the data emphasizes multi-step reasoning and logical inference. According to the company, each dataset includes structured problem-solving processes and step-by-step reasoning paths, enabling AI models to generate more stable and explainable outputs.

The datasets are built in standardized formats and categorized by difficulty and task type, with multi-stage quality assurance processes in place. This allows client organizations to immediately apply the data to model training and evaluation while significantly reducing the time and cost required for data design and preprocessing.

A company representative stated that EBIT possesses the capability to expand beyond mathematics and coding into industry and professional domains such as finance, business, science and technology, and general knowledge. The company also offers customized datasets tailored to specific needs, including financial calculations, risk analysis, quantitative decision-making, and domain-specific reasoning.

Leveraging its experience in AI data construction projects for global technology companies, EBIT has recently expanded collaborations with domestic LLM developers and AI startups. The company noted that its experience in building multilingual, high-complexity reasoning datasets has enabled it to accumulate global-level data quality standards and operational expertise.

CEO Yonghan Park said, “We aim to help SMEs, startups, and research institutions efficiently utilize high-quality mathematics and coding-based AI training data,” adding, “By delivering scalable data construction capabilities across diverse industry domains, we will support our clients in enhancing AI model performance and strengthening their R&D competitiveness.”