IATP — AI-driven reverse engineering and autonomous development pipeline

The challenge

IATP is a complex legacy platform that had evolved over years through multiple development teams, with minimal documentation and no standardised architecture. The original developers were no longer available. The client needed to understand what the system actually did at a functional and architectural level, produce comprehensive analysis documentation, and then modernise and extend the platform with a reliable, repeatable development process.

Manual reverse engineering at this scale would have taken months and produced documents that were outdated before completion. A traditional rewrite was too risky without first having a verified understanding of existing business logic.

Phase 1 — AI-powered reverse engineering

We deployed Nexus MDS Core as the analytical backbone. The entire legacy codebase was ingested into a RAG pipeline powered by Weaviate, creating a semantically searchable knowledge base of every module, function, data flow, and business rule embedded in the code.

Using multi-agent workflows orchestrated through n8n, the system automatically generated context maps — reconstructing the functional architecture, identifying hidden dependencies, mapping data lineage, and flagging undocumented business logic. This context was then used to auto-generate structured analysis documents: functional specifications, architectural diagrams descriptions, data dictionaries, and risk assessments — all produced via intelligent RAG queries against the ingested codebase.

As the analysis progressed and new documents were generated, the platform's AI-driven obsolescence engine automatically tracked which earlier analysis artifacts were superseded by newer, more complete versions — maintaining an obsolescence graph so the development team always worked from the most current understanding of the system.

What would have taken a team of analysts months was produced in days, with traceability back to the source code for every claim in every document.

Phase 2 — AI-driven development pipeline

With the system fully understood and documented, we implemented a modern, AI-augmented development pipeline:

• Antigravity — AI-assisted code generation and refactoring, using the RAG-generated context as the grounding layer so every code change was informed by verified understanding of the existing system.
• Trae IDE — integrated development environment with AI copilot capabilities, connected to the project's knowledge base for context-aware code completion and inline documentation.
• MCP Servers — Model Context Protocol servers for DevOps orchestration: automated deployment pipelines, infrastructure provisioning, environment management, and CI/CD integration — all AI-coordinated.
• Comet — AI-driven test generation and quality assurance. Test cases automatically derived from the functional specifications produced in Phase 1, with continuous regression testing and coverage analysis to ensure modernisation did not break existing behaviour.

Results

The project delivered a complete, end-to-end AI-driven software lifecycle: from understanding a legacy system with no documentation, through automated analysis document generation, to a modern development pipeline where AI assists at every stage — coding, testing, deployment, and quality assurance.

The RAG-powered reverse engineering reduced the analysis phase from an estimated 3-4 months to under two weeks. The AI-driven development pipeline increased development velocity while maintaining quality through automated, context-aware testing. The entire approach is now repeatable and forms part of the Dynamics Consulting methodology for legacy modernisation engagements.

Technologies used