Recruiting · Production

LinkedIn builds a multi-step ML pipeline to extract and map skills from content into the LinkedIn Skills Graph

The problem

Skills embedded in member profiles, job postings, and learning course descriptions are often absent from dedicated skills sections, making comprehensive and accurate mapping to the LinkedIn Skills Graph difficult.

Workflow diagram · grounded in source
1
Profile or content update triggers extraction
trigger
“LinkedIn's standardized member profile skills feature requires nearline inference when a member profile is created or updated”
2
Skill segmentation
ai_action
“we first parse the raw input into a well-formed structure. A job posting, for example, may have sections for "company description," "responsibilities," "benefits," and "qualifications." Meanwhile, a resume will usually have sections that…”
3
Skill tagging
ai_action
“a skill tagger identifies the mentions of skills in the text. The skill tagger can perform token-based skill matches and also perform semantic-based skill matches from short sentences/phrases”
4
Skill expansion via Skills Graph
ai_action
“The skill expansion relies on our Skills Graph to query for relevant skills in the same skill group or skills that share structural relationships, such as parent skills, children skills, and sibling skills.”
5
Multitask cross-domain scoring
ai_action
“a multitask scoring runs to identify each content piece and skill candidate pair. The multitask scoring model contains two parts, a shared module and a domain-specific module”
6
Serve skills to downstream systems
output
“including search, recommendations, feed ranking, Jobs You May Be Interested In, Job Search, Recruiter Search, and many others”
7
Feedback loops for model improvement
feedback_loop
“we build feedback loops directly into online Job Posting and member profiles to help with AI model iterations”
Reported outcome

The multitask learning framework for job-important skills yielded improvements across job recommendation, job search, and job-member skill matching, with Knowledge Distillation reducing the model size by 80% without compromising performance.

Reported metrics
Model size reduction via Knowledge Distillation80%
Job Recommendation: Member Job Applicants and Offsite Apply Clickers+0.1391%
Job Recommendation: Predicted Confirmed Hires+0.4606%
Job Search: Job Sessions+0.1468%
Show all 10 reported metrics
Model size reduction via Knowledge Distillation80%
Job Recommendation: Member Job Applicants and Offsite Apply Clickers+0.1391%
Job Recommendation: Predicted Confirmed Hires+0.4606%
Job Search: Job Sessions+0.1468%
Job Search: PPC revenue+0.7577%
Job Search: Engagements+0.2271%
Job Member Skills matching: Qualified Applications+0.87%
Job Member Skills matching: Qualified Application Rate+0.40%
Job Member Skills matching: Predicted Confirmed Hires+0.24%
Job Member Skills matching: Applicants and Apply Click Counts+0.48%
Reported stack
LinkedIn Skills GraphMultilingual BERTTransformerKnowledge DistillationSamza-BEAMSpark
Source
https://engineering.linkedin.com/blog/2023/extracting-skills-from-content-to-fuel-the-linkedin-skills-graph
Read source ↗

Frequently asked questions

What did this team achieve with this AI workflow?

The multitask learning framework for job-important skills yielded improvements across job recommendation, job search, and job-member skill matching, with Knowledge Distillation reducing the model size by 80% without c…

What tools did this team use?

LinkedIn Skills Graph, Multilingual BERT, Transformer, Knowledge Distillation, Samza-BEAM, Spark.

What results were reported?

Model size reduction via Knowledge Distillation: 80%; Job Recommendation: Member Job Applicants and Offsite Apply Clickers: +0.1391%; Job Recommendation: Predicted Confirmed Hires: +0.4606%; Job Search: Job Sessions: +0.1468% (source-reported, not independently verified).

How is this recruiting AI workflow structured?

Profile or content update triggers extraction → Skill segmentation → Skill tagging → Skill expansion via Skills Graph → Multitask cross-domain scoring → Serve skills to downstream systems → Feedback loops for model improvement.

Related recruiting cases