The mission of LinkedIn is to create opportunity for every member of the global workforce. One way we realize this mission is connect members and employers. Many of our Talent Solutions products are built around mathematical models that try answer a simple question: is this opportunity of interest to this member at this time? Answering that question requires a multidisciplinary approach, drawing on tools from machine learning and data mining and utilizing insights from psychological and sociological research. 

Our models are trained on labeled data sets. We employ professional recruiters to tag member-job pairs for us (good fit, not a good fit). For positive labels, we also use our members current positions as well as their job transitions. 

A big challenge is that member preferences change over time. At certain points in their tenure, people are likely to look for a new job. People are sometimes willing to relocate to a new city, but at other times they want to stay in the same place. Many people are happy with their job, and will only move for a promotion. Other people may be dissatisfied and interested in doing the same job at a competing firm. We use long term and short term personalization methods to capture these latent preferences

An interesting feature space for our models is the Person-Environment Fit (P-E fit). It captures the degree to which a person fits within their workplace. Tools from machine learning and data science applied to the LinkedIn Economic Graph can help us quantify certain aspects of P-E fit and be predictive of workplace success and happiness. In particular, Person-Job fit measure how well a person has the skills and experience necessary for a specific role. A particular challenge with matching members and jobs is that both member profiles and job advertisements are incomplete. In order to be successful, we must infer the missing information. For instance, we can utilize similar profiles and endorsements to learn inferred skills for each member. We can utilize career trajectory to infer likely next steps. We can use the characteristics of members who apply to a job to create a virtual profile for that position.