Hanmi Pharmaceutical’s 4,000 Quality Documents to Be Analyzed by AI

[Financial News] Megazone Cloud is building an AI agent that will search and analyze more than 4,000 quality documents from Hanmi Pharmaceutical and automatically compare them with the latest regulations. The goal is to automate quality assurance (QA) tasks currently performed manually and reduce working time by more than 40%.
Megazone Cloud announced on the 7th that it had begun building Hanmi Q Agent, an AI agent designed to automate Hanmi Pharmaceutical’s QA operations. The project is part of the AI Transformation (AX) One-Stop Voucher Support Program led by the Ministry of Science and ICT and the National IT Industry Promotion Agency (NIPA).
Hanmi Q Agent is a service that searches and analyzes more than 4,000 quality documents accumulated by Hanmi Pharmaceutical. It applies retrieval-augmented generation (RAG) technology based on Megazone Cloud’s proprietary enterprise AI operating system, AIR Studio. The system is designed to search and summarize quality documents, compare them with the latest regulations, and identify sections that do not comply with newly added or revised requirements.
The AI response-generation engine will use Solar, a domestically developed large language model (LLM) from Upstage. Megazone Cloud plans to fine-tune Solar for Hanmi Pharmaceutical’s good manufacturing practice (GMP) domain before integrating it into an AIR Studio-based agent.
Quality documents are materials prepared and managed to demonstrate that medicines have been manufactured consistently in accordance with regulations. Because they contain tables, figures, and formulas, as well as a mixture of Korean and English, employees have had to locate the necessary documents themselves and compare them individually with the latest regulations and internal guidelines. Noncompliant sections have also generally been managed through corrections made after inspections.
Once Hanmi Q Agent is deployed, the latest regulations can be continuously compared with the quality documents. This is expected to shift quality management from a reactive approach—correcting problems after inspections—to a preventive approach that identifies potential regulatory noncompliance in advance.
Megazone Cloud Vice President Seo Mintaek said, "Having validated our AIR Studio and fine-tuning capabilities through generative AI projects across various industries, we will use this project to lead the expansion of domain-specific AI in regulated industries."


[email protected] Choi Hyerim Reporter