From Traditional Machine Learning to Large Language Models: The Evolution of Infrastructure from an Engineering Perspective

The transition to large language models (LLMs) represents a major shift from traditional machine learning (ML) modeling. it raises new requirements for the technological systems supporting model development and deployment. Organizations that have built robust healthcare systems with traditional ML are now dealing with the need to support LLMs, which require large amounts of data, different preprocessing techniques, and new forms of deployment. Based on the authors’ experience in building and operating both traditional ML systems and LLM systems, this paper analyzes the evolution of AI infrastructure required to support LLMs from an engineering perspective. It covers systematic changes in data processing, model versioning, resource allocation, and monitoring across different system stages. The paper recommends infrastructure that supports both systems, and describes which components can be retained and which should be upgraded during the transition. I hope to provide guidance to engineering teams considering an upgrade to their AI infrastructure while preserving their existing systems.