Knowledge-grounded AI engineering workflows

Enhancement and modernisation of a legacy enterprise product: extending backend capabilities, adding web-based UI support and integrating a REST and gRPC microservices architecture that serves both WPF and web clients, including SAP integration.

Alongside core .NET development, I was selected as an AI Champion within the engagement, applying GenAI and agentic workflows to design, implementation and legacy modernisation.

Highlights

• Built a knowledge-grounding layer for AI agents: RAG over user manuals and project information, a code-derived knowledge graph across ~37 repositories and ~40M lines of code, repository-specific code intelligence and curated Markdown knowledge
• Integrated retrieval into AI workflows through MCP-based interfaces so agents can reach project context and tools
• Designed AI-assisted workflows with specialised sub-agents and orchestration across design, analysis and implementation
• Contributed strategy and workflow design for modernising a legacy regression-testing tool — an approach targeting a multi-year effort in under six months, subject to validation of the remaining milestones
• Resolved long-standing legacy defects dating back to roughly 2009–2012 while preserving backward compatibility