Skip to main content Skip to navigation

AI-powered Code Review with LLMs: Early Results

Project Overview

The document explores the transformative role of generative AI, particularly through Large Language Models (LLMs), in education, focusing on a novel AI-powered code review system designed to enhance software quality and developer learning. This innovative system identifies code issues, suggests optimizations, and streamlines the code review process by delivering actionable feedback, thereby addressing the shortcomings of conventional static analysis tools. The findings emphasize the efficacy of LLMs in improving educational outcomes for developers and revolutionizing software development practices. By leveraging generative AI, the initiative not only supports learners in acquiring better coding skills but also fosters a more dynamic and responsive learning environment that adapts to individual needs, showcasing the broader potential of AI technologies in reshaping educational methodologies and enhancing the quality of software engineering.

Key Applications

LLM-based AI agent model for code reviews

Context: Software development lifecycle targeting developers

Implementation: Trained on large datasets of code repositories, including code reviews and bug reports, utilizing GitHub APIs for data access.

Outcomes: Enhanced code quality, reduced post-release bugs, and improved developer education through actionable feedback.

Challenges: Integration of AI into existing workflows and ensuring accuracy in suggestions.

Implementation Barriers

Technical Barrier

Existing traditional tools lack the depth to provide actionable feedback beyond basic syntax error detection.

Proposed Solutions: Developing LLM-based models that can understand context and provide detailed suggestions for code improvement.

Integration Barrier

The challenge of integrating AI solutions into established software development workflows.

Proposed Solutions: Demonstrating the effectiveness of AI tools to stakeholders and gradually incorporating them into the development process.

Project Team

Zeeshan Rasheed

Researcher

Malik Abdul Sami

Researcher

Muhammad Waseem

Researcher

Kai-Kristian Kemell

Researcher

Xiaofeng Wang

Researcher

Anh Nguyen

Researcher

Kari Systä

Researcher

Pekka Abrahamsson

Researcher

Contact Information

For information about the paper, please contact the authors.

Authors: Zeeshan Rasheed, Malik Abdul Sami, Muhammad Waseem, Kai-Kristian Kemell, Xiaofeng Wang, Anh Nguyen, Kari Systä, Pekka Abrahamsson

Source Publication: View Original PaperLink opens in a new window

Project Contact: Dr. Jianhua Yang

LLM Model Version: gpt-4o-mini-2024-07-18

Analysis Provider: Openai

Let us know you agree to cookies