Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

LinkSAGE: Optimizing Job Matching Using Graph Neural Networks

Ping Liu, Haichao Wei, Xiaochen Hou, Jianqiang Shen, Shihai He, Kay Qianqi Shen, Zhujun Chen, Fedor Borisyuk, Daniel Hewlett, Liang Wu, Srikant Veeraraghavan, Alex Tsun, Chengming Jiang, Wenjing Zhang

TL;DR

LinkSAGE tackles the problem of scalable, dynamic job matching on LinkedIn by embedding a large, heterogeneous job marketplace graph within an inductive GNN framework and coupling it to existing DNN ranking models via an encoder–decoder architecture. The solution employs a nearline inference pipeline to precompute GNN embeddings and feed them into downstream models with low latency, avoiding costly real-time GNN computation. Across multiple online AB tests, LinkSAGE achieves consistent improvements in engagement, relevance, and retention, and it enhances equity for users with sparse historical data. The work demonstrates a practical, production-ready pathway for deploying GNNs in industrial-scale recommender systems and highlights the GNN-based platform as a reusable signal-engineering layer across the Job Marketplace stack.

Abstract

We present LinkSAGE, an innovative framework that integrates Graph Neural Networks (GNNs) into large-scale personalized job matching systems, designed to address the complex dynamics of LinkedIns extensive professional network. Our approach capitalizes on a novel job marketplace graph, the largest and most intricate of its kind in industry, with billions of nodes and edges. This graph is not merely extensive but also richly detailed, encompassing member and job nodes along with key attributes, thus creating an expansive and interwoven network. A key innovation in LinkSAGE is its training and serving methodology, which effectively combines inductive graph learning on a heterogeneous, evolving graph with an encoder-decoder GNN model. This methodology decouples the training of the GNN model from that of existing Deep Neural Nets (DNN) models, eliminating the need for frequent GNN retraining while maintaining up-to-date graph signals in near realtime, allowing for the effective integration of GNN insights through transfer learning. The subsequent nearline inference system serves the GNN encoder within a real-world setting, significantly reducing online latency and obviating the need for costly real-time GNN infrastructure. Validated across multiple online A/B tests in diverse product scenarios, LinkSAGE demonstrates marked improvements in member engagement, relevance matching, and member retention, confirming its generalizability and practical impact.

LinkSAGE: Optimizing Job Matching Using Graph Neural Networks

TL;DR

LinkSAGE tackles the problem of scalable, dynamic job matching on LinkedIn by embedding a large, heterogeneous job marketplace graph within an inductive GNN framework and coupling it to existing DNN ranking models via an encoder–decoder architecture. The solution employs a nearline inference pipeline to precompute GNN embeddings and feed them into downstream models with low latency, avoiding costly real-time GNN computation. Across multiple online AB tests, LinkSAGE achieves consistent improvements in engagement, relevance, and retention, and it enhances equity for users with sparse historical data. The work demonstrates a practical, production-ready pathway for deploying GNNs in industrial-scale recommender systems and highlights the GNN-based platform as a reusable signal-engineering layer across the Job Marketplace stack.

Abstract

We present LinkSAGE, an innovative framework that integrates Graph Neural Networks (GNNs) into large-scale personalized job matching systems, designed to address the complex dynamics of LinkedIns extensive professional network. Our approach capitalizes on a novel job marketplace graph, the largest and most intricate of its kind in industry, with billions of nodes and edges. This graph is not merely extensive but also richly detailed, encompassing member and job nodes along with key attributes, thus creating an expansive and interwoven network. A key innovation in LinkSAGE is its training and serving methodology, which effectively combines inductive graph learning on a heterogeneous, evolving graph with an encoder-decoder GNN model. This methodology decouples the training of the GNN model from that of existing Deep Neural Nets (DNN) models, eliminating the need for frequent GNN retraining while maintaining up-to-date graph signals in near realtime, allowing for the effective integration of GNN insights through transfer learning. The subsequent nearline inference system serves the GNN encoder within a real-world setting, significantly reducing online latency and obviating the need for costly real-time GNN infrastructure. Validated across multiple online A/B tests in diverse product scenarios, LinkSAGE demonstrates marked improvements in member engagement, relevance matching, and member retention, confirming its generalizability and practical impact.
Paper Structure (19 sections, 6 equations, 4 figures, 10 tables)

This paper contains 19 sections, 6 equations, 4 figures, 10 tables.

Table of Contents

  1. Introduction
  2. Related Work
  3. Graph Construction
  4. Model Training
  5. Data Preparation
  6. Model Architecture
  7. Training and Optimization
  8. Model Serving
  9. Integration with Ranking Models
  10. Nearline Inference Framework
  11. GNN as a Platform for Feature Engineering
  12. Experiments
  13. Top Applicant Jobs (TAJ)
  14. Job You May Be Interested In (JYMBII)
  15. Job Search
  16. ...and 4 more sections

Figures (4)

  • Figure 1: An example of LinkedIn's job marketplace graph. Members (also referred to as job seekers) are connected to their attributes, including Skill, Title, Company, and Position. The same connection is applied for each job. A member is connected to a job as well if there are engagements happening (e.g., save, apply, or click).
  • Figure 2: Model Architecture
  • Figure 3: Model training and serving.
  • Figure 4: The working flow of nearline inference pipeline for job embeddings.