AWS Amarathon 2025 Recap

Accelerating Migration Projects with Kiro using Spec-Driven Development

Recap Series

Session Notes

What is Amazon Kiro?

  • Compatible with VS Code
  • Utilizes cutting-edge Claude models
  • Enterprise-grade security Closing the Idea-to-Code Gap
  • Addresses the challenge of product ideas losing fidelity when passed from PM to engineer
  • Kiro acts as an AI co-author, converting informal briefs into living, version-controlled specifications
  • Eliminates ambiguity before coding starts, accelerating delivery cycles Traditional Static Docs vs. Kiro's Living Specs
  • Traditional static docs are prone to staleness and lead to high rework and surprises
  • Kiro's Living Specs auto-update with code, sync with tests & metrics, and result in measurable gains in velocity Amazon Kiro Workflow
  • SDLC in Amazon Kiro Way!
  • Ingest: Product brief intake
  • Expand: AI expands user stories
  • Iterate: Collaborative editing
  • Lock & Stub: Spec finalization
  • Every step is traceable, commentable, and under Git control

Prompt Engineering for Precision

  • Concise, context-rich prompts yield the most accurate and useful specifications Fill-in-the-Blank Template:
  • Persona: Who is the user?
  • Problem: What do they need to do?
  • Outcome: What does success look like?
  • Non-Goals: What is out of scope? Iterative Refinement Process:
  • Human Prompts
  • AI Suggests
  • Human Validates
  • This cycle repeats until the spec passes all internal quality gates

Modernizing Legacy with Amazon Kiro

  • Prescriptive pattern to migrate critical COBOL, PL/I, and Assembler to manage Java on Graviton without rewriting core business logic
  • Automated Transformation: Converts legacy code to modern Java microservices
  • Data Replication: Syncs mainframe data to cloud databases in real-time
  • DevOps Integration: Establishes CI/CD pipelines for rapid, reliable releases Automated Code Transformation Engine
  • Parse: Legacy source is converted into an Abstract Syntax Tree (AST)
  • Refactor: Rule-based transformations are applied to modernize the code structure
  • Emit: Idiomatic Java code is generated as Spring Boot microservices
  • Preserves data formats, transaction boundaries, and audit trails Data Sync & Mainframe Off-Loading
  • AWS DMS and Kiro agents replicate mainframe data to Aurora PostgreSQL in near-real time
  • Bidirectional Sync: Mainframe remains authoritative during pilot phases
  • Zero-Downtime Cut-Over: Traffic is switched to the cloud via a simple DNS flip Zero-Trust Security
  • Every microservice is isolated and secured by default
  • Least-Privilege IAM: Unique IAM role with minimal permissions
  • Encryption: KMS encrypts data at rest and in transit
  • Continuous Monitoring: CloudTrail and Guard Duty provide audit evidence for SOX, PCI, and HIPAA Role-Based Access & Reviews:
  • [ 1 ] Product Manager: Owns the narrative and user-facing requirements
  • [ 2 ] Tech Lead: Owns the system architecture and technical design
  • [ 3 ] QA Engineer: Owns the acceptance criteria and test plans
  • Amazon Kiro tracks approvals, surfaces unresolved comments, and blocks merge until all roles sign off Real-World Impact Case Study: Global Bank
  • A top-20 bank used Kiro to migrate 14 million lines of COBOL for retail payments
  • Seamless cut-over over a single weekend with zero failed transactions
  • 68% cost reduction in infrastructure
  • 90% reduction in MIPS (Mainframe load)
  • Release frequency rose from quarterly to weekly, freeing budget for AI initiatives Your Migration Roadmap
  • 6-Week Pilot Blueprint:
  • Week 1-2: Provision & Transform
  • Week 3: Validate Parity
  • Week 4: CI/CD Pipeline
  • Week 5: 5% Traffic
  • Week 6: Metrics & Review