A Scalable and Efficient Signal Integration System for Job Matching

2025-07-13

AI Summary

LinkedIn’s recruitment recommendation system faces three key challenges: cold-start issues, information silos, and systemic matching bias. To address these, we propose STAR—a novel collaborative modeling framework integrating large language models (LLMs) and graph neural networks (GNNs). LLMs enhance semantic understanding to mitigate cold-start problems; GNNs encode professional relationship graphs to break information silos; and joint optimization alleviates structural bias in job-candidate matching. STAR further introduces adaptive negative sampling and embedding versioning to support multi-source signal fusion and end-to-end industrial-scale embedding deployment. Evaluated on LinkedIn’s large-scale production traffic, STAR significantly improves both matching accuracy and diversity. It has been fully deployed across LinkedIn’s core recommendation services. The system advances both algorithmic innovation—through principled LLM-GNN co-design—and engineering scalability—via robust infrastructure for real-time, high-throughput embedding serving.

Technology Category

• Machine Learning: Large Multimodal Models (LMMs)
• Data Mining & Knowledge Management: Conversational Systems for Recommendation & Retrieval
• Natural Language Processing: (Large) Language Models

Application Category

• Graph Algorithms and Modeling for the Web: Foundation models and LLMs for Web-related graphs
• User Modeling, Personalization and Recommendation: Large Language Models (LLM) for user modeling and recommendation
• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking

Abstract

LinkedIn, one of the world's largest platforms for professional networking and job seeking, encounters various modeling challenges in building recommendation systems for its job matching product, including cold-start, filter bubbles, and biases affecting candidate-job matching. To address these, we developed the STAR (Signal Integration for Talent And Recruiters) system, leveraging the combined strengths of Large Language Models (LLMs) and Graph Neural Networks (GNNs). LLMs excel at understanding textual data, such as member profiles and job postings, while GNNs capture intricate relationships and mitigate cold-start issues through network effects. STAR integrates diverse signals by uniting LLM and GNN capabilities with industrial-scale paradigms including adaptive sampling and version management. It provides an end-to-end solution for developing and deploying embeddings in large-scale recommender systems. Our key contributions include a robust methodology for building embeddings in industrial applications, a scalable GNN-LLM integration for high-performing recommendations, and practical insights for real-world model deployment.

Problem

• Addressing cold-start and bias in job matching recommendations
• Integrating LLMs and GNNs for scalable signal processing
• Developing end-to-end embeddings for industrial recommender systems

Innovation

• Combines LLMs and GNNs for signal integration
• Uses adaptive sampling and version management
• Provides end-to-end embedding deployment solution