This research introduces a novel approach using Large Language Models enhanced with contrastive learning to automatically decompose monolithic applications into microservices.

• Leverages contrastive learning techniques to improve LLM performance on code-based tasks
• Automates the complex and costly process of identifying service boundaries
• Generates more coherent and maintainable microservice architectures compared to traditional approaches
• Addresses technical debt while enabling more scalable and flexible systems

For engineering teams, this technology promises to significantly reduce the time and resources required for modernizing legacy applications, potentially transforming how organizations approach digital transformation initiatives.

Contrastive Learning-Enhanced Large Language Models for Monolith-to-Microservice Decomposition