Automated Code quality assessment and LLM driven Refactoring methadology

This project presents an intelligent system that checks the quality of computer code and improves it automatically. First, the system evaluates the code using different software quality factors such as complexity, errors, duplication, security issues, and maintainability. These factors are combined to calculate a final quality score to check if the code is good enough or not. If the code does not meet the required standard, an AI model (LLM) is used to improve and rewrite the code while keeping its original purpose the same. After improvement, another AI model (CodeBERT) checks whether the new code still has the same meaning as the original code. This is done by comparing both codes using similarity measurement. The system only accepts improvements when the code is both better in quality and still works the same. It also learns from past results to improve its decision-making over time. Overall, this project helps in automatically improving code quality in a smart, reliable, and efficient way.

Poster