2025 Volume 27 Issue 1 Pages 73-76
This report explores the potential applications and challenges of generative AI in the pharmaceutical industry from the perspective of its users. It begins by outlining the fundamental concepts of prompt engineering and Retrieval-Augmented Generation (RAG) tuning, presenting methods for optimizing AI-generated outputs. Additionally, using the generative AI application development platform “Dify,” a demonstration was conducted based on the FDA’s concept of Computer Software Assurance (CSA), in which the effects of prompt design and the presence or absence of RAG tuning on AI responses were compared and analyzed. The results suggest that prompt engineering and knowledge enhancement through RAG tuning are effective for generating contextually appropriate responses. On the other hand, challenges such as accuracy, reproducibility, and ethical risks were also identified. To ensure the appropriate use of generative AI, it is important not only to promote user understanding of the technology, but also to design operational workflows that incorporate the Human-in-the-Loop approach.
