Augmenting large language models with static code analysis for accelerated software development and quality improvements

This thesis presents the Next-Generation Quality Accelerator (NGQA), an automated end-to-end pipeline driven by large language models (LLMs) to accelerate the quality assurance (QA) phase of the software development lifecycle using static analysis tools. NGQA integrates detection, grounding, revision, validation, and coordination into a unified workflow through a verification-aware artificial intelligence (AI) agent architecture. The pipeline consists of six steps that combine SonarQube for static issue detection, retrieval-augmented generation (RAG) based false positive mitigation, LLM-based code revision, structural dependency analysis, test case generation using Local Chain-of-Thought (LCoT) reasoning, and multi-metric quality validation. It supports both cloud-based and fully local LLM deployment. Empirical evaluation on 70 repositories across seven programming languages shows that NGQA resolves 83.5% of issues, improves PassRatio by 16.5%, CodeBLEU by 28.8%, CodeScore by 24.0%, and achieves 89% F1-score in false positive mitigation compared to baseline methods. NGQA achieves a 32.6-fold acceleration over manual QA, completing the evaluation in 95.9 hours compared to an estimated 3,126 person-hours manually.

Name thesis:degree_name

Master of Applied Science (MASc)

Discipline thesis:degree_discipline

Software Engineering

Grantor

University of Ontario Institute of Technology

Year dc:date.issued

2026

Author dc:creator

• Abtahi, Seyed Moein

Advisor dc:contributor.advisor

• Azim, Akramul

Language dc:language.iso

en

Handle dc:identifier.uri

https://hdl.handle.net/10155/2073

OAI identifier oai:identifier

oai:ontariotechu.scholaris.ca:10155/2073