Building The Next Generation Of Job Search At Linkedin

LinkedIn's natural-language job search separates firm requirements from broader preferences. Its query engine interprets a request, retrieves relevant profile or network information, and converts strict criteria into filters. Other criteria become a numerical representation used to find jobs with related meaning.

For example, the company describes resolving a requested metropolitan area and identifying employers where the job seeker already has connections. Those become explicit search constraints. Smaller models then retrieve and rank listings, trained to follow a larger model's relevance judgments. This lets a job seeker describe an intention without supplying every exact keyword, while retaining filters for requirements that should not be treated loosely.