Abstract: Techniques are disclosed for generating instruction sets and initializing executable system resources using dynamic graph data structures. A system detects a system-generated event signal based on a machine-readable data stream associated with a user data structure, where the data stream integrates first metadata from internal logged interactions and second metadata from external system events stored on remote repositories. The system constructs a computer-generated graph data structure with vertices representing entries from both metadata sources, and establishes linkages reflecting their relationships. A model queries the graph data structure, analyzing event signals, vertices, and linkages to produce an enhanced feature set. An instruction set, including structured data fields initialized with the enhanced feature set, is generated and associated with the user data structure.