AI-Based Code Generation and Analysis: A Paradigm Shift in Software Development

Rupendra Jaiswal*, Shakshi Srivastava**
* Department of Computer Science and Engineering (M. Tech), S.R. Institute of Management & Technology, Lucknow, Dr. APJ Abdul Kalam Technical University, Uttar Pradesh, India.
** Department of Computer Science and Engineering, S.R. Institute of Management & Technology, Lucknow, Dr. APJ Abdul Kalam Technical University, Uttar Pradesh, India.
Periodicity:July - September'2025

Abstract

Artificial Intelligence (AI) is transforming software engineering by enabling automated code generation, intelligent debugging, and quality assurance workflows. This paper presents CodeSynth+, a hybrid framework that integrates transformer based code generation with Graph Neural Network (GNN)-based semantic analysis to produce syntactically correct, semantically validated, and secure multi-language code (Python, Java, C++). CodeSynth+ operates in an iterative feedback loop: natural language requirements are converted to initial code by a fine- tuned transformer, the code is parsed into Abstract Syntax Trees (ASTs) and semantic graphs, and a GNN inspects structural and data-flow properties to detect logic errors and vulnerabilities. We describe dataset construction (public code corpora, competitive programming solutions, and curated GitHub projects), formalize evaluation metrics (Syntactic Accuracy, Semantic Precision, Maintainability Index, Security Vulnerability Score, Regeneration Success Rate, and Time to Production), and detail baseline configurations (CodeT5, Codex-like evaluation, and SonarQube rules). Ten experiments demonstrate consistent improvements versus transformer only baselines: increased semantic accuracy and vulnerability detection, improved maintainability scores, and reduced time-to-production. We provide statistical validation (multiple seeds, means ± std, confidence intervals, and significance testing) and reproducibility artifacts (scripts and configs) in the supplemental repository. CodeSynth+ represents a step toward autonomous, interpretable, and secure code generation in modern software engineering.

Keywords

Artificial Intelligence, Graph Neural Network, Abstract Syntax Tree, Structured Query Language, Static Analysis Static Code Analysis, Vulnerability Detection.

How to Cite this Article?

Jaiswal, R., and Srivastava, S. (2025). AI-Based Code Generation and Analysis: A Paradigm Shift in Software Development. i-manager’s Journal on Software Engineering, 20(1), 37-43.

References

[3]. Chen, M. (2021). Evaluating large language models trained on code. arXiv preprint arXiv:2107.03374.
[9]. Vaswani, A., Shazeer, N., Parmar, N., Uszkoreit, J., Jones, L., Gomez, A. N., & Polosukhin, I. (2017). Attention is all you need. Advances in Neural Information Processing Systems, 30.
If you have access to this article please login to view the article or kindly login to purchase the article

Purchase Instant Access

Single Article

North Americas,UK,
Middle East,Europe
India Rest of world
USD EUR INR USD-ROW
Pdf 35 35 200 20
Online 15 15 200 15
Pdf & Online 35 35 400 25

Options for accessing this content:
  • If you would like institutional access to this content, please recommend the title to your librarian.
    Library Recommendation Form
  • If you already have i-manager's user account: Login above and proceed to purchase the article.
  • New Users: Please register, then proceed to purchase the article.