A twenty-step transformation lifecycle now enables the airline to upgrade outdated Java frameworks and resolve hard-coded version numbers with minimal manual intervention from developers. This initiative addresses a substantial legacy Java platform comprising more than 250 business services and roughly 750 individual code repositories. By utilizing AWS Transform Custom, the engineering teams are now effectively moving away from aging infrastructure and adopting a streamlined, containerized environment that supports growth while reducing technical debt across systems.
1. Phase 1: Detailed Deterministic Preparation
The initial stage of the transformation focuses on establishing a comprehensive understanding of the current service landscape through automated data collection and baseline analysis. GitHub Actions workflows are deployed to scan each service repository, identifying the specific production-versioned artifacts that are currently in use. Simultaneously, the system extracts traffic patterns by analyzing live production logs, determining both mean and peak usage rates for every service. This data is vital for calculating non-functional requirements such as resource sizing, scaling thresholds, and observability settings.
Following the initial analysis, the preparation phase moves into establishing performance benchmarks and provisioning the target infrastructure for each service. Automated testing packs are executed against the existing platform to create a known-good baseline for both regression and performance metrics. These results serve as a mandatory standard for comparison once the transformation is complete, ensuring that the new environment performs as well as the legacy system. Once the benchmarks are established, the system automatically provisions the required infrastructure in the target architecture, creating a clean environment for the modernized code.
2. Phase 2: Executing Agentic Code Transformation
In the second phase, the focus shifts to the modification of the codebase, where agentic AI takes the lead in performing complex updates across the repository. The engine automatically upgrades the service code to Java 21, ensuring that the airline is utilizing the latest features and security patches provided by the language. Beyond the language upgrade, the AI migrates deployment patterns from traditional Amazon EC2 configurations to Amazon ECS on AWS Fargate. This includes optimizing the code to run efficiently on AWS Graviton processors, which requires specific adjustments to take full advantage of the ARM-based architecture.
A critical aspect of the agentic transformation involves the elimination of hard-coded version numbers and other static configurations that hinder operational flexibility. The AI engine systematically identifies these blockers and replaces them with dynamic references, allowing for easier maintenance and more frequent security updates. This level of intricate logic change would normally consume months of engineering time per service if done manually, but the AI completes it with remarkable speed and accuracy. Once the core transformation is finished, the engineering team performs final repository checks to audit the code for any remaining hard-coded elements or anomalies.
3. Phase 3: Comprehensive Testing and Deployment
The final phase of the modernization lifecycle ensures that the updated services are fully functional and ready for production use through a series of rigorous tests. The service then undergoes the same automated regression and performance testing that was performed during the preparation phase. By comparing these new results against the established known-good benchmarks, the system can verify that the business logic has been preserved and that performance meets standards. This validation step is essential for maintaining trust in the automated process, as it provides concrete evidence that the AI-driven changes have not introduced any regressions.
Once the automated tests are successful, the service enters a human-in-the-loop review process to ensure it meets the airline’s high standards for quality and security. Professional engineers conduct manual code reviews to verify that the AI-generated changes align with internal best practices and architectural guidelines. Additionally, the service is subjected to intensive security and code quality scans to identify any potential vulnerabilities that might have been introduced during the transformation. Final approval must be obtained from cybersecurity and operations teams before the service is officially cleared for go-live.
4. Strategic Outcomes and Future Considerations
The implementation of AWS Transform Custom has yielded substantial strategic benefits, most notably in the increased speed of modernization across the Java estate. By automating the most time-consuming aspects of code refactoring and dependency management, the airline has significantly reduced the manual effort required from its development teams. Furthermore, the shift to a containerized, auto-scaling architecture on Amazon ECS provides enhanced scalability that is essential for handling the volatile demand patterns of the travel industry. This improved operational efficiency not only reduces costs but also provides a more stable platform for critical systems. The successful modernization of hundreds of code repositories proved that combining agentic AI with deterministic testing was the most effective way to handle legacy debt. This project established a new benchmark for technical transformations within the airline industry, demonstrating that even the most complex business services could be updated with minimal disruption. For organizations looking to replicate this success, the primary takeaway was the importance of establishing a 20-step lifecycle that prioritizes automated validation. The focus should remain on creating repeatable, auditable processes that leverage machine intelligence to maintain a secure and modern software estate while fostering continuous innovation.
