LinkedIn's Hiring Assistant uses a plan-and-execute architecture to automate repetitive recruiting tasks while keeping humans in control of key decisions. The system splits work between a Planner that breaks down requests into steps and an Executor that completes them. A supervisor agent coordinates specialized sub-agents for intake, sourcing, evaluation, outreach, screening, learning, and memory. The assistant operates in interactive mode for initial setup and asynchronous mode for large-scale candidate searches across 1.2 billion profiles. Built on a message-driven, event-driven architecture with GraphQL APIs and agent-driven UI, it integrates LinkedIn's Economic Graph for market intelligence. Quality is maintained through policy rails and human alignment, with custom AI models optimized for speed and accuracy in candidate evaluation.
Table of contents
Solve Enterprise Auth, Identity, and Security for Your App (Sponsored)Better Deals, By Design with Verizon This Holiday (Sponsored)The High-Level ArchitectureThe Agentic User ExperienceTechnical Building BlocksThe Supervisor AgentThe Specialized Sub-AgentsThe Quality PillarsConclusionSPONSOR USSort: