AWS re:Invent 2025 - Modernizing Java applications with generative AI (DVT210)

Modernizing Java Applications with Generative AI

Challenges with Java Application Modernization

• Developers often inherit legacy Java applications written 10+ years ago, before cloud computing
• Upgrading these applications poses several "modernization monsters":
  • Documentation Dragon: Extensive documentation updates required when upgrading Java versions
  • Testing Troll: Newer Java versions require different testing approaches, breaking existing unit tests
  • Module Monster: Upgrading one part of the application can negatively impact interconnected modules
  • Version Vampire: Breaking changes between Java versions, like the removal of JAXB in Java 11
  • Dependency Dinosaur: Compatibility issues with third-party libraries and dependencies
  • Time-Sucking Tarantula: The overall process can take months or years, often getting deprioritized

Solving Modernization Challenges with AI

• AWS introduced the Java Transformation Agent, an AI-powered tool to automate the modernization process:
  • Verifies the project, analyzes dependencies, and generates a custom transformation plan
  • Performs code generation and verification builds in the cloud, with local unit test execution
  • Provides a file diff to review changes before accepting them
  • Allows chunking large projects into smaller, less risky upgrade patches

ADP's Java Modernization Journey

• ADP is a global leader in payroll and HR services, serving 1.1 million customers and 42 million workers
• Faced challenges with Java 8 and 11 end-of-life, compliance requirements, security vulnerabilities, and performance
• Partnered with AWS to pilot the Java Transformation Agent on their global payroll portal, which had 40+ Java services
  • Able to transform the application in 3-4 hours instead of 3-4 days, a 60% time savings
  • Improved application resiliency, reliability, and performance by leveraging the latest Java features
  • Automated unit test case updates and developed new functional test automation scripts

AWS Transform Custom: The Next Generation

• Limitations of the current Java Transformation Agent:
  • Only handles runtime upgrades, not deployment or UI layer upgrades
  • IDE-based, not integrated with CI/CD pipelines or custom build systems
  • Lacks flexibility for targeted upgrades or organization-specific requirements
• Introducing AWS Transform Custom:
  • A command-line interface-based autonomous agent that can learn from organizational context
  • Supports any code patterns, not just Java, and various scenarios like language conversions or architectural changes
  • Provides a continuous learning capability, where the agent improves based on user feedback
  • Can be integrated into CI/CD pipelines and scaled to upgrade thousands of applications

Scaling Java Modernization with AWS Transform Custom

• AWS Transform Custom can be wrapped in a batch script to execute transformations at scale across multiple repositories
• Demonstrated a batch script that ran the Java 8 to 21 upgrade in parallel across several GitHub projects
  • Achieved a 60% success rate, with the ability to review and enable knowledge items for future runs
• The centralized transformation definition and knowledge base allow organizations to:
  • Empower developers with reusable, vetted transformation capabilities
  • Establish quality gates and change management processes for modernization initiatives
  • Accelerate Java upgrades by 5x compared to manual efforts

Key Takeaways

• AWS provides a comprehensive solution to address the "modernization monsters" plaguing Java application upgrades
• The Java Transformation Agent and AWS Transform Custom leverage AI and generative capabilities to automate the modernization process
• Customers like ADP have achieved 60% faster upgrades and improved application quality, reliability, and performance
• AWS Transform Custom offers a flexible, scalable, and continuously learning approach to modernize Java and other applications