Stephen Chin and Neeraj Deshmukh explain why converting a single formula can miss surrounding jobs, data layouts and transaction behavior. They propose using a knowledge graph to expose relationshipsA knowledge graph represents entities and their relationships in a structured form that can be queried and connected. and provide relevant context to agentsContext engineering designs how relevant instructions, evidence, state and tool information are assembled for a model's task..
Neeraj Deshmukh describes agent task plans, human approval and validation across generated code and runtime behaviorEvaluation is the systematic assessment of a system's behaviour or results against explicit criteria.. An interest calculation illustrates why rounding and state changes require comparison against the original system.
The speakers present customer examples and reported turnaround times as their companies’ own claims, including a 35-million-line codebase mapped in four days. The talk does not provide an independent evaluation of those results.
Watch on YouTube
