Modern cloud migrations require analysing vast volumes of source code, documentation, and architectural metadata, tasks that exceed the limits of manual review and often obscure critical dependencies. This talk presents a practical blueprint for building agentic AI systems that enable scalable semantic analysis and automated reasoning across heterogeneous enterprise artefacts.

We will explore how unstructured data is transformed into queryable knowledge through embedding pipelines, including ingestion strategies, metadata lineage, chunking design, and vector indexing. Particular focus is given to how embedding granularity, retrieval accuracy, and latency trade offs directly impact downstream reasoning quality and system performance.

Building on this foundation, the talk introduces a multi agent orchestration layer where specialised agents collaborate to perform complex tasks such as dependency mapping, cross document reasoning, task decomposition, and migration strategy synthesis. We will examine coordination patterns, guardrails, and failure handling mechanisms required to ensure robustness in production environments. Attendees will gain practical insights into designing scalable embedding pipelines, aligning retrieval architecture with reasoning objectives, and implementing reliable multi agent workflows.