AWS re:Invent 2025 - Best practices for performing custom code transformation with agentic AI-MAM344

AWS re:Invent 2025 - Best Practices for Performing Custom Code Transformation with Agentic AI

Overview

• Presentation on the challenge of technical debt and scaled remediation options
• Introduction to AWS Transform Custom, a new tool for performing custom code transformations at scale

The Technical Debt Challenge

• Significant cost and time required to address technical debt
• Slows down innovation and prevents delivery of new features
• Introduces security vulnerabilities and maintenance burdens
• Leads to performance limitations and strategic misalignment

Existing Approaches and Limitations

• Rule-based automation (e.g. grep, open rewrite) can be brittle and require specialized expertise
• General-purpose coding AI tools (e.g. Curo, Codex) lack consistency across developers and require human-in-the-loop
• Learnings remain siloed, leading to repeated discovery of the same issues

AWS Transform Custom

• Designed to discover and learn custom code transformation patterns
• Allows teaching the agent your own transformation logic
• Executes transformations at scale in a headless, automated fashion
• Continuously learns and improves based on feedback and execution history

Key Features

• CLI interface for interactive and non-interactive (batch) execution
• Transformation definition registry for sharing and collaborating
• Integrated planning, execution, and validation with self-debugging
• Continuous learning to improve transformation quality over time

Transformation Definition Process

1. Expert defines initial transformation using interactive CLI
2. Transformation is refined through iterative execution and feedback
3. Transformation definition is published to a central registry
4. Teams can pull and execute the transformation at scale
5. Feedback and learnings are incorporated back into the definition

AWS-Provided Transformations

• Out-of-the-box transformations for common use cases:
  • Java, Python, Node.js runtime upgrades
  • AWS SDK version upgrades
  • Early access: Comprehensive codebase analysis, x86 to Graviton migration
• Validated and benchmarked by AWS, available for immediate use

Best Practices

• Start with pilots and small-scale testing to evaluate efficacy and cost
• Leverage existing workflows and tooling, minimizing disruption
• Utilize the AWS Transform web application for campaign management and reporting

Business Impact

• Enables scaled remediation of technical debt across an organization
• Accelerates modernization efforts and reduces manual effort
• Promotes consistency and shared learnings across teams
• Unlocks the ability to tackle complex, custom transformation challenges

Example Use Cases

• Air Canada: Upgrading thousands of deprecated Lambda functions
• Twitch: Migrating 900+ applications from AWS SDK v1 to v2
• QAD: Automating migration of customer-specific ERP customizations
• MongoDB: Upgrading Java codebase from Java 8/11 to Java 21+

Conclusion

AWS Transform Custom provides a powerful, scalable, and customizable solution for addressing technical debt through automated code transformations. By leveraging agentic AI and a collaborative transformation definition process, organizations can unlock significant productivity gains, improve code quality, and accelerate their modernization efforts.