Recruiting · Production

LinkedIn extends GenAI tech stack to build multi-agent AI systems including Hiring Assistant

The problem

LinkedIn's existing GenAI stack supported single-agent, one-off prompt interactions but was not tenable for the complex, long-running multi-agent workflows that members and customers needed across the platform.

Workflow diagram · grounded in source
1
Agent definition and registration
integration
“Developers simply annotate this definition with some platform defined proto3 options that describe the metadata of their agent, and register it via a build plugin into the skill registry (a central service that tracks available agents, t…”
2
Multi-agent orchestration via messaging
routing
“our messaging system best emulated the characteristics we wanted in a multi-agent orchestrator. Long lived tasks could be broken down into a sequence of messages with guaranteed first in first out (FIFO) delivery and seamless message his…”
3
Human-in-the-loop control
human_review
“agents seek clarification, get feedback, or request approvals at key decision points, balancing autonomy with control”
4
Context engineering and RAG
ai_action
“RAG and knowledge graphs surface the latent meaning hidden within data, making it more usable for both agents and users. Central to this transformation is context engineering—a practice that involves feeding LLMs with the right data and …”
5
Agent response delivery
output
“It allows agents to respond in a single chunk, incrementally across a single synchronous response (synchronous streaming) or even split their responses across multiple asynchronous messages (asynchronous streaming); thus allowing us to m…”
6
Trace-based continuous improvement
feedback_loop
“Execution traces are persisted and aggregated into datasets that power offline evaluations, model regression tests, and prompt tuning experiments. These traces become the raw material for understanding agent behavior over time, detecting…”
Reported outcome

LinkedIn extended its GenAI platform with multi-agent orchestration, human-in-the-loop control, and layered observability, and is making Hiring Assistant globally available in English to customers.

Reported metrics
Hiring Assistant global availabilityglobally available in English to customers
Sync delivery speed improvementsignificantly speeds up delivery with predictable times
Reported stack
gRPCLangSmithLangGraphLangChainOpenTelemetryMCPA2A
Source
https://www.linkedin.com/blog/engineering/generative-ai/the-linkedin-generative-ai-application-tech-stack-extending-to-build-ai-agents
Read source ↗

Frequently asked questions

What did this team achieve with this AI workflow?

LinkedIn extended its GenAI platform with multi-agent orchestration, human-in-the-loop control, and layered observability, and is making Hiring Assistant globally available in English to customers.

What tools did this team use?

gRPC, LangSmith, LangGraph, LangChain, OpenTelemetry, MCP, A2A.

What results were reported?

Hiring Assistant global availability: globally available in English to customers; Sync delivery speed improvement: significantly speeds up delivery with predictable times (source-reported, not independently verified).

How is this recruiting AI workflow structured?

Agent definition and registration → Multi-agent orchestration via messaging → Human-in-the-loop control → Context engineering and RAG → Agent response delivery → Trace-based continuous improvement.

Related recruiting cases