Empowering Application Modernization with LLMs: Addressing Core Challenges in Reliability, Security, and Quality
Ahilan Ayyachamy Nadar Ponnusamy · 2025
Read Empowering Application Modernization with LLMs: Addressing Core Challenges in Reliability, Security, and Quality by Ahilan Ayyachamy Nadar Ponnusamy (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2506-10984.txt.
Introduction
Ahilan Ayyachamy Nadar Ponnusamy, a solo author, proposes a framework for updating old software with large language models, in this 2025 paper. AI code tools write fast, but their output can carry security holes and uneven quality. The framework leans on two model strengths, reasoning about code and generating code, and pairs both with human guidance at each step.
A detailed case study walks through one real modernization from start to finish. At each stage, the paper weighs other approaches and shows why the chosen path won. A working reference version is shared on GitHub, so teams can adapt the framework to their own legacy systems.
Catalog record
| CREATOR | Ahilan Ayyachamy Nadar Ponnusamy |
|---|---|
| YEAR | 2025 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 26 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2506-10984
Browse the full library · About the Tower of Babel
Full text
The opening of the work, from the library's hosted text:
Empowering Application Modernization with LLMs: Addressing Core Challenges in Reliability, Security, and Quality Ahilan Ayyachamy Nadar Ponnusamy Abstract AI-assisted code generation tools have revolutionized software development, offering unprecedented efficiency and scalability. However, multiple studies have consistently highlighted challenges such as security vulnerabilities, reliability issues, and inconsistencies in the generated code. Addressing these concerns is crucial to unlocking the full potential of this transformative technology. While advancements in foundational and code specialized language models have made notable progress in mitigating some of these issues, significant gaps remain, particularly in ensuring high quality, trustworthy outputs. This paper builds upon existing research on leveraging large language models (LLMs) for application modernization. It explores an opinionated approach that emphasizes two core capabilities of LLMs: code reasoning and code generation . The proposed framework integrates these capabilities with human expertise to tackle application modernization challenges effectively. It highlights the indispensable role of human involvement and guidance in ensuring the success of AI-assisted processes. To demonstrate the framework’s utility, this paper presents a detailed case study, walking through its application in a real-world scenario. The analysis includes a step-by-step breakdown, assessing alternative approaches where applicable. This work aims to provide actionable insights and a robust foundation for future research in AI-driven application modernization. The reference implementation created for this paper is available here GitHub Repository . Index Terms —Artificial Intelligence, Application Modernization, Large Language Models, Legacy Application, Code Documentation, Code Generation, Application Modernization. 1 INTRODUCTION Legacy systems, particularly those developed during the dot-com boom and early phases of digital transformation, represent a critical challenge for modern enterprises. While these systems have been the operational backbone for a couple of decades, they are increasingly hampered by technological obsolescence, scalability issues, and maintenance complexity. Within the scope of this paper, we focus on legacy systems that utilize relatively modern programming paradigms and languages, excluding older systems developed in languages like COBOL and Fortran. The exclusion of older languages is driven primarily by the limited public knowledge and training data available for them, which reduces the effectiveness of large language models (LLMs) in reasoning about or generating code for these technologies. Instead, this paper focuses on modernization efforts for systems developed in more contemporary languages, such as Java and .NET, hereafter referred to as legacy applications. These systems are well suited for leveraging LLMs, given the abundance of available training data and the demonstrated proficiency of LLMs with recent technologies. Additionally, as these applications often operate across the internet, mobile, and edge environments, they face heightened security and compliance risks, making their timely modernization crucial. Code specialized, open source LLMs have matured to a point where they can meaningfully contribute to application modernization without requiring costly and time intensive fine-tuning. This is particularly important for lowering barriers to entry, a key focus of this paper. By leveraging readily available, open source models, organizations can explore modernization with minimal investment in infrastructure or specialized expertise. Another aspect of this paper is the emphasis on accessibility and democratization. The proposed framework is built for enterprise-scale deployment. The reference implementation tool, while optimized to run efficiently on modest hardware like a MacBook Pro, can seamlessly scale for enterprise-level deployments. This makes it feasible for a wide range of users to experiment, customize, and adopt the framework based on their specific needs. The key contributions of this paper are as follows: • Addressing LLM Limitations for Legacy Systems: The paper explores the challenges faced by LLMs in modernizing legacy applications, particularly when direct code translation approaches are employed [ 1 ] . • Leveraging Code Reasoning for Modernization: By utilizing the improving code reasoning capabilities of LLMs, the framework enables thoughtful and context-aware modernization processes [ 2 ] . • Adopting “Progressive Prompting” for Incremental Progress: An effective technique, “progressive prompting,” is introduced to streamline the process of generating, refining, and validating modernized code in incremental steps [ 3 ] . • Integrating Security Best Practices and IP Protection: To mitigate risks and ensure intellectual property (IP) protection, the framework incorporates security-focused best practices, making it enterprise-ready [ 4 ] . • Democratizing Adoption: The paper proposes a standardized framework that reduces the complexity of experimentation and adoption. This is particularly relevant for enterprises seeking to explore modernization without incurring significant upfront costs. • Promoting Local Deployment: By leveraging open models, the paper highlights a lightweight, local deployment approach that minimizes the need for specialized hardware and cloud services, lowering the barrier to experimentation and early adoption. This paper builds on existing research to present a pragmatic and accessible approach to AI-assisted application modernization. By focusing on a framework that integrates LLM capabilities with human oversight, it bridges the gap between theoretical potential and practical implementation, offering actionable insights for enterprises seeking to modernize legacy systems. 2 RESEARCH METHOD To address the challenges and complexities outlined above, this paper adopts a structured approach divided into two key subcategories: 2.1 Requirements Generation for Legacy Applications The first step involves leveraging LLMs to analyze legacy application code and generate comprehensive functional requirements. These requirements are then reviewed and refined by SMEs to ensure accuracy, relevance, and alignment with modernization objectives. This step serves as the foundation for building a modernized application that aligns with both current and future business needs. 2.2 Development of the New Application Based on Refined Requirements Using the reviewed and refined requirements, the framework guides the generation of a new application in a step-by-step manner. This process focuses on translating the modernized requirements into functional code, utilizing the reasoning and code generation capabilities of LLMs. The overall process is supported by a robust framework and a purpose-built reference implementation tool designed to standardize and streamline the modernization workflow. This framework provides clear guidance for each phase of the process, ensuring consistency, repeatability, and the integration of human oversight at critical decision points. By organizing the research into these two subcategories and grounding the process in a practical framework and toolset, this paper provides a comprehensive, accessible, and actionable approach to AI-assisted application modernization. 3 THE FRAMEWORK AND THE TOOLl The framework aims to simplify and streamline the modernization of legacy applications by providing a step-by-step approach, from generating requirements to producing modern, enhanced code for the target platform. It focuses on two key areas: • Requirements Generation from Legacy Applications • Development of a New Application Based on Refined Requirements 3.1 Requirements Generation from Legacy Applications The following interaction diagram illustrates the requirements generation from legacy applications process in detail. Figure 1: Interaction diagram showing the requirements generation process. The framework recommends starting with the Interaction Layer , which serves as the entry point for users (via UI) and other systems (via APIs or Batch scripts). Since this layer typically captures most of the functional requirements, it forms a solid starting point. Next, attention shifts to the Business Logic Layer , where specific calculations and business rules are implemented. Finally, the process moves to the Data Layer or Configuration Files , which store system level information, properties, and lists. By covering these three layers, the framework ensures that the core functional requirements of the legacy application are captured, providing a strong foundation for modernization. Figure 2: Recommended steps for Requirements Generation.