Transforming Career Search Experience with AI

Careers are deeply personal journeys—shaped by pivotal moments, hard-earned growth, unexpected turns, and the pursuit of meaning. For many, a job isn’t just a paycheck—it’s a path to purpose. But navigating that path can be overwhelming. From searching for the right role, to figuring out if it’s a true fit, to waiting (often endlessly) to hear back—each step in the journey carries emotional weight. And when the system falls short, it doesn’t just slow people down—it leaves them stuck.

LinkedIn research finds that despite high motivation, 60% of professionals who started the year wanting a new job and over 240 million on LinkedIn who have signaled they’re open to work, both seekers and hirers are frustrated. Forty percent of job seekers are applying less or giving up entirely, while another 40% are casting a wider net just to improve their odds. On the other side of the marketplace, hirers are frustrated with managing the sheer volume of unqualified applications. The tools we use today take a one-size-fits-all approach to a highly personal process. Fixing this requires a complete reimagining—using AI to build a more adaptive, human-centered experience. 

Tailored discovery that understands you

That reimagination starts with the first step: job search. Finding the right role is often the biggest hurdle—especially when you don’t know what keywords to use, what titles to search for, or how to express what you really want. That’s why we’ve built the new search experience that lets members describe their goals in their own words and get results that truly reflect what they’re looking for—even roles they didn’t know existed. This is the first step in a larger journey to make job-seeking more intuitive, inclusive, and empowering for everyone.

While traditional job search tools are bounded by predefined filters and keywords, AI Job Search more closely mirrors a discussion with a career advisor or a trusted colleague. Queries aren’t confined by rigid categories and members can use conversational language, for instance: 

• “I want to find entry level jobs in video games.”
• “Use marketing skills to cure cancer”
• “I want to make cities more walkable and bike-friendly”

Integrating natural language into the search engine

AI Job Search is built on large language models (LLMs) that have been fine-tuned with rich and unique insights from LinkedIn’s Economic Graph. 

This is the first time LinkedIn has applied LLMs across the entire stack of our search and recommender systems from understanding a member’s search query to retrieving all possible jobs and finally ranking the most relevant jobs first. These models power the search with a deeper understanding of the intent, phrasing, and nuances behind the natural language queries. To deliver the best relevancy, we use a cross-encoder LLM to score and rank jobs with the highest precision. 

To make this cutting-edge approach both fast and scalable, we developed an LLM distillation technique. This compresses the model while preserving quality, enabling real-time job ranking with lower computing cost by pre-processing key information from both the job opportunity and the search query when possible and avoiding starting from scratch each time. 

This complex modeling wouldn’t be possible without the efficiency of GPU-powered computing. But our journey didn’t start overnight. Two years ago, during a routine team meeting, two engineers floated an ambitious and somewhat improbable idea: using GPUs to facilitate extensive retrieval capabilities. Rather than dismissing it, we quickly tested the concept. When the results proved promising, we pivoted the technology to support our Job Search product. Today, our system can scan the entire corpus of job listings in milliseconds—like flipping through a library of books to find the exact one you need. And, it’s one of the world’s largest libraries of jobs!

Robust rubric and evaluation

To ensure we surface the right jobs to each member, we complement our LLM-powered search engine with structured evaluation guidelines. These guidelines, called the relevance rubrics, are created by our job experts guided by the product policy. This defines what “relevant” means in the context of each pair of query and job. We then use LLMs to apply the various relevancy rubrics and evaluate our performance automatically, helping us surface jobs that are better aligned with our member’s goals, skills, and experiences. 

AI Job Search is just one step in our broader ambition to support every job seeker more personally and effectively. The team is constantly testing and learning with this technology and as we continue exploring what’s possible with AI, we’ll create even more customized experiences to support members as they pursue their dream career.