The Sequence Pulse: The ML Architecture Powering LinkedIn's Skills Graph
Using transformer models to map jobs to job seekers.
The Sequence Pulse is the segment where we discuss how AI practitioners are building solutions in the real world. Today, we are going to deep dive into one of the most interesting AI solutions built internally at LinkedIn.
Throughout the last few years, LinkedIn has built one of the most innovative machine learning(ML) infrastructures in the world. Many of the technologies and architectures built at LinkedIn has served as reference and inspiration for ML open source frameworks and, more importantly, include patterns that can be used by technologists building ML solutions in their companies. Today, we are going to deep dive into another LinkedIn ML architecture. This time we will focus on skill mapping.
Skills mapping for job searches is at the center of LinkedIn’s value proposition. In the professional landscape, the value of skills cannot be overstated. Over the years, LinkedIn evolves its skill mapping technology into a sophisticated stack known as the LinkedIn Skills Graph. The goal of the LinkedIn Skill Graph is to facilitate job searches, skill acquisition, and opportunity evaluation. The aim is to build a vast, reliable, and accurate skills repository.
To ensure a robust skills-based framework, LinkedIn integrates all skills on its platform with the Skills Graph. While skills listed on profiles or job descriptions are easily mapped, the challenge lies in identifying skills within LinkedIn Learning courses, resumes, and posts.
Let’s dive in….