Recruiting · Production

Building LinkedIn's Hiring Assistant: an agentic AI for recruiter sourcing, evaluation, and engagement at global scale

The problem

Candidate sourcing and evaluation are the most resource-intensive parts of recruiting, combining high-value decision-making with large volumes of repetitive pattern-recognition tasks that require human micromanagement to keep pipelines updated.

First attempt

A simple ReAct-style architecture was found insufficient for enterprise-grade recruiting: LLMs did not follow instructions reliably, produced hallucinations, and complex reasoning introduced unacceptable latency.

Workflow diagram · grounded in source
1
Recruiter initiates request
trigger
“The interactive interface allows recruiters to converse directly with Hiring Assistant. This helps clarify hiring requirements, align on expectations, and adjust behavior before large-scale execution begins.”
2
Intake: gather role requirements
ai_action
“Gathers and refines hiring requirements and role details from recruiters. It confirms key attributes like job title, location, and seniority, inferring missing information when needed. It then generates role-specific qualifications from …”
3
Sourcing: generate and run queries
ai_action
“Generates multiple search queries based on hiring requirements and evaluation criteria. Leveraging and enhancing LinkedIn Recruiter Search and Recommended Matches tools, it runs these queries at scale, stores potential candidate profiles…”
4
Evaluation: assess candidates
ai_action
“Assesses candidates by synthesizing information from multiple sources (including profiles, resumes, and historical engagement data) and applies hiring requirements and evaluation rubrics to produce structured recommendations with reasoni…”
5
Human review of recommendations
human_review
“Recruiters remain in the loop to review these insights and make the final decisions on advancing candidates through the recruiting workflow.”
6
Outreach: contact candidates
output
“Handles communication with candidates, including generating and sending initial outreach and follow-up messages across multiple channels. It also replies to candidate questions based on the hiring requirements and FAQs defined during the…”
7
Learning: refine from recruiter signals
feedback_loop
“Continuously refines hiring requirement personalization by analyzing recruiter actions, such as adding candidates to pipelines or sending InMails. It updates qualifications and candidate recommendations dynamically by integrating both ex…”
Reported outcome

Hiring Assistant reached global availability, enabling sourcing across 1.2+ billion profiles with enterprise-grade throughput and evaluating candidates in seconds via custom fine-tuned LLMs, freeing recruiters to focus on high-value work.

Reported metrics
Candidate profiles searchable1.2+ billion profiles
Candidate evaluation speedevaluate candidates in seconds
Pipeline refresh frequencyRefreshes candidate pipelines daily
Reported stack
LinkedIn Recruiter SearchRecommended MatchesEconomic GraphLinkedIn Talent InsightsGraphQL API
Source
https://www.linkedin.com/blog/engineering/ai/how-we-engineered-linkedins-hiring-assistant
Read source ↗

Frequently asked questions

What did this team achieve with this AI workflow?

Hiring Assistant reached global availability, enabling sourcing across 1.2+ billion profiles with enterprise-grade throughput and evaluating candidates in seconds via custom fine-tuned LLMs, freeing recruiters to focu…

What tools did this team use?

LinkedIn Recruiter Search, Recommended Matches, Economic Graph, LinkedIn Talent Insights, GraphQL API.

What results were reported?

Candidate profiles searchable: 1.2+ billion profiles; Candidate evaluation speed: evaluate candidates in seconds; Pipeline refresh frequency: Refreshes candidate pipelines daily (source-reported, not independently verified).

What failed first in this deployment?

A simple ReAct-style architecture was found insufficient for enterprise-grade recruiting: LLMs did not follow instructions reliably, produced hallucinations, and complex reasoning introduced unacceptable latency.

How is this recruiting AI workflow structured?

Recruiter initiates request → Intake: gather role requirements → Sourcing: generate and run queries → Evaluation: assess candidates → Human review of recommendations → Outreach: contact candidates → Learning: refine from recruiter signals.

Related recruiting cases