Artificial intelligence is becoming an increasingly important area of software engineering, influencing how developers approach software design, coding, testing, deployment, and maintenance. Machine learning, deep learning, natural language processing, and expert systems are being explored for tasks ranging from requirements analysis to defect detection and project planning.
As these technologies become part of software development workflows, an important question emerges: how can AI be integrated into the Software Development Life Cycle (SDLC) while maintaining software reliability, security, and quality?
Mahathi Kari is working on this area of software engineering, examining AI methodologies, their applications across the SDLC, and the technical considerations involved in adopting intelligent systems for software development. Her work looks at how AI can support different stages of the development process while also considering limitations related to accuracy, privacy, bias, and responsible implementation.
Understanding AI Methodologies in Software Development
Artificial intelligence encompasses several methodologies that serve different purposes in software engineering. Machine learning can analyze historical information to identify patterns relevant to defect analysis, project planning, and resource allocation. The reliability of these applications depends on the quality of the data and its relevance to the problem being addressed.
Natural language processing enables systems to interpret and generate human language. Within software development, it can assist with requirements analysis, technical documentation, stakeholder communication, and identifying relationships between written specifications and other development artifacts.
Deep learning and neural networks can support pattern recognition and selected code-related tasks. Expert systems apply predefined rules to structured decisions, while reinforcement learning explores decision-making through feedback. Understanding the differences between these approaches is important when determining which methodology may be suitable for a particular software engineering problem.
Mahathi’s examination of these methodologies considers their place within the broader development process, where the objective is not simply to introduce automation but to understand how intelligent techniques can be applied to specific engineering tasks.
Applying AI Across the Software Development Life Cycle
The Software Development Life Cycle provides a structured process for planning, designing, building, testing, deploying, and maintaining software. AI can support activities throughout these stages, although its role depends on the project’s requirements and the level of automation appropriate to the task.
During planning, AI-assisted analysis can help examine project information and identify potential scheduling or resource constraints. During requirements analysis, natural language processing can assist with organizing stakeholder feedback and reviewing written specifications. At the design stage, AI tools can help developers explore architectural alternatives and identify possible inconsistencies.
During implementation, coding assistants can suggest code snippets, explain existing implementations, and support debugging. In testing, AI-assisted systems can help generate test cases and analyze potential defects. After deployment, monitoring tools can examine operational information and flag unusual behavior for further investigation.
These applications form an important part of the area Mahathi is examining: how different AI methodologies relate to individual SDLC stages and what considerations arise when they are incorporated into software engineering workflows.
AI-Assisted Coding and Code Optimization
AI-assisted coding has become a prominent application of artificial intelligence in software development. Tools such as GitHub Copilot and ChatGPT can help developers interpret programming instructions, explore implementation options, generate code suggestions, and understand existing code.
AI can also assist with code review and refactoring by identifying duplicated logic, complex structures, or potentially inefficient algorithms. These suggestions may help developers examine maintainability and consider alternative implementations.
However, generated code can misunderstand requirements, contain functional errors, or introduce security vulnerabilities. Developers need to review the output, test the implementation, and check compatibility with the surrounding codebase before incorporating changes into a software system.
This distinction between generating code and verifying its correctness is an important consideration when examining AI’s role in software engineering. Automation can support implementation, but engineering judgment remains necessary to determine whether the resulting software behaves as intended.
AI in Software Testing and Quality Assurance
Software testing evaluates whether an application behaves according to its requirements. AI-assisted testing tools can help generate test cases, identify possible edge cases, and examine patterns in historical defect information.
Machine learning can also be used to analyze code changes and identify components that may warrant additional testing. Such predictions can inform testing priorities, but they do not establish that a defect exists or guarantee that unflagged components are error-free.
Automating repetitive testing activities can allow engineers to focus on complex scenarios and exploratory testing. The effectiveness of these techniques depends on test quality, evaluation criteria, and the interpretation of results.
Mahathi’s focus on AI applications across the SDLC also brings attention to the need to consider the limitations of these tools alongside their potential uses. In quality assurance, this means distinguishing between identifying a possible problem and establishing through appropriate testing that the software meets its requirements.
AI in Project Planning and Resource Management
Software project management involves coordinating schedules, resources, technical dependencies, and changing requirements. AI-assisted analysis can examine historical project information and identify patterns associated with delivery timelines, resource needs, and potential risks.
In agile development environments, AI tools can assist with organizing project information, summarizing task updates, and examining development trends. These capabilities can provide additional context for planning discussions and help teams evaluate competing priorities.
Project forecasts nevertheless depend on assumptions and available data. Unexpected technical problems, changes in project scope, and resource constraints can affect delivery. AI-generated estimates should therefore be treated as decision-support information rather than guaranteed predictions.
Examining project management alongside coding and testing provides a broader view of AI in software engineering. Its applications extend beyond writing code to include the planning and coordination activities that influence how software is developed and maintained.
Requirements Traceability and Impact Analysis
Requirements traceability connects a software system’s requirements with its design, implementation, and testing activities. Maintaining these relationships can become more difficult as projects expand and specifications change.
Natural language processing can help identify related information across requirements documents, design specifications, source code, and test cases. AI-assisted analysis may also help developers investigate which software components could be affected by a proposed change.
These capabilities can support impact analysis and help identify possible gaps in documentation. However, automatically identified relationships require verification because similar terminology does not necessarily mean that two technical artifacts have the same meaning.
For AI-assisted traceability to be useful, the relationships identified by a system must be evaluated in the context of the actual software requirements. Clear specifications and human review remain essential to maintaining reliable connections between development activities.
AI in Deployment and Software Maintenance
Deployment and maintenance involve preparing applications for release, monitoring their behavior, addressing defects, and adapting systems to changing requirements. AI-assisted monitoring can analyze system logs, operational metrics, and events to identify unusual patterns that may require investigation.
Predictive approaches can use historical and current system information to help examine potential failures. Within DevOps workflows, automation may also support deployment checks, incident analysis, and selected recovery procedures.
Production environments require careful controls because automated changes can affect availability, system behavior, and data integrity. Testing, monitoring, access restrictions, and appropriate approval procedures help ensure that automated operations remain consistent with operational requirements.
Considering these stages alongside development and testing helps explain why AI in software engineering must be assessed across the entire lifecycle rather than through an isolated task.
Artificial Intelligence in Software as a Medical Device
AI also has applications in specialized software environments, including software as a medical device, commonly known as SaMD. Depending on its intended purpose, such software may support activities related to diagnosis, monitoring, or treatment.
Machine learning and other AI techniques can be applied to medical information to identify patterns and support selected clinical workflows. These applications require attention to data quality, reliability, validation, and privacy because software behavior may influence clinical decisions.
AI-based medical software also raises regulatory considerations. Systems that learn or change over time can present additional challenges for evaluation and ongoing oversight. The presence of AI does not, by itself, establish that a medical software product is safe or effective. Appropriate evidence and applicable regulatory requirements remain important.
For software engineers, these applications illustrate why the context in which an AI system operates matters. The controls required for a medical application may differ considerably from those appropriate for a general-purpose development assistant.
Reliability, Privacy, and Responsible AI Governance
AI integration introduces challenges that require attention throughout development and deployment. Reliability is one concern because generated code and automated recommendations may be incorrect. Models trained on incomplete or biased information can produce unsuitable suggestions, and their behavior may vary when they encounter unfamiliar inputs.
Data privacy and security are equally important. Development workflows may involve proprietary source code, credentials, customer information, and sensitive system logs. Organizations need to understand how their selected tools process information, restrict access appropriately, and follow relevant security and data-protection requirements.
Bias and transparency present further considerations. AI systems may reflect limitations in their training data or produce recommendations that are difficult to explain. Overreliance on automated suggestions can also reduce the scrutiny applied to generated output.
AI governance provides a framework for addressing these concerns through defined responsibilities, validation procedures, documentation, monitoring, and review. The appropriate controls depend on the intended application, the sensitivity of the data, and the consequences of potential errors.
These issues form an important part of Mahathi Kari’s examination of AI in software engineering, which considers not only where intelligent technologies may be applied but also the limitations that need to be considered when integrating them into development processes.
Research Directions in AI-Driven Software Engineering
Research in AI-driven software engineering examines how intelligent techniques can be incorporated into established development practices. Areas of interest include project risk management, data mining, software testing, computational intelligence, and governance across the SDLC.
Mahathi’s work explores these themes in relation to the different stages of software development. This includes examining how AI methodologies may support planning, implementation, testing, deployment, and maintenance, alongside the challenges that arise when automated systems are used to support engineering decisions.
Other approaches, including federated learning and edge AI, may be relevant to applications with specific data-handling, computing, or architectural requirements. Their suitability depends on the problem being addressed and the environment in which the software must operate.
Continued examination of these areas can help clarify where AI-assisted methods are appropriate, what limitations need to be considered, and how their outputs should be evaluated within established software engineering practices.
A More Considered Approach to AI in Software Engineering
Artificial intelligence offers software engineering teams additional ways to analyze requirements, assist with coding, support testing, examine project information, and monitor deployed applications. Its potential applications extend throughout the Software Development Life Cycle, but the usefulness of each approach depends on the task, the quality of its output, and the controls surrounding its use.
Mahathi Kari’s work in this area centers on understanding AI methodologies and their applications across software engineering, while considering the practical and ethical questions associated with their adoption. This includes examining how AI can support different development activities without overlooking reliability, privacy, bias, and the importance of human oversight.
As intelligent tools continue to become part of software development, evaluating their role requires a balanced approach. Combining AI-assisted capabilities with systematic testing, engineering judgment, and appropriate governance provides a framework for exploring their applications while retaining responsibility for the software systems being developed and maintained.
