Leveraging Large Language Models in Software Development Lifecycle - Banani Roy



AI Summary

Summary of Dr. Bon Roy’s Talk on Leveraging Large Language Models in Software Development Lifecycle

  1. Introduction to Dr. Bon Roy
    • Assistant Professor, University of Saskatchewan
    • Director of the Interactive Software Engineering and Analytics Lab
    • Recognized for contributions to software development and mentoring.
  2. Research Focus
    • Investigation into human-centric software development.
    • Issues caused by software bugs and legacy software due to developer mistakes.
  3. The Role of AI in Software Development
    • Acknowledged developer challenges in software maintenance.
    • Importance of using AI tools (especially large language models (LLMs)) to reduce bugs and migration issues.
  4. Large Language Models (LLMs)
    • Overview of capabilities in software engineering, including:
      • Requirement analysis
      • Code generation and completion
      • Bug detection
      • Test case generation.
  5. Research Contributions
    • User Story Generation: Introduced a new prompt engineering technique called Refine and Thought (RAT) for generating user stories from requirement documents.
    • Autogenics: Developed a method for generating inline comments for code snippets using LLMs to enhance code comprehension.
  6. Recognition of Effectiveness
    • Presented results from evaluations indicating improvements in user story and inline comment quality.
    • Highlighted the need for further studies to expand on findings and increase dataset comprehensiveness.
  7. Conclusion and Future Work
    • Emphasized the necessity for fine-tuned LLMs to prevent hallucinations and improve performance in software engineering tasks.
    • Plans to continue exploring AI’s role in enhancing developer productivity and code quality.
  8. Questions & Engagement
    • Open floor for audience questions regarding LLM methodologies and developer engagement in software projects.