Recently, LinkedIn officially launched a brand-new AI-powered job search feature for all users. This new function leverages refined and fine-tuned large models with the aim of improving the job-seeking experience on the platform. Erran Berger, Vice President of Product Development at LinkedIn, stated: "This new search experience allows users to describe their job goals in their own words, thereby obtaining more relevant results."

AI robot interviews and negotiations

Image source note: Image generated by AI, image authorized service provider Midjourney

Previously, when users searched for jobs on LinkedIn, they often relied on precise keyword queries, which led to many job seekers receiving job recommendations that did not match their true needs. For example, after entering "reporter," users might see different types of job recommendations such as media reporters and court reporters, both of which require completely different skills. Wenjing Zhang, Vice President of Engineering at LinkedIn, pointed out that to better help users find suitable jobs, the platform needs to improve its understanding of user queries.

The new system allows users to search using natural language. For instance, they can input "find recent software engineer positions released in Silicon Valley" instead of just simple job titles. This change makes LinkedIn's understanding of user needs more intelligent.

To achieve this goal, LinkedIn comprehensively revamped its search function. First, they needed to enhance the system's ability to understand query content, then retrieve relevant information from the vast job database, and finally rank the found jobs to ensure the most relevant ones appear first. Previously, LinkedIn used fixed classification methods and older ranking models, lacking deep semantic understanding, so they turned to modern large language models (LLMs) to improve natural language processing capabilities.

However, using large models is also computationally expensive. To reduce costs, LinkedIn adopted a model distillation method, breaking the LLM into two steps: one for data retrieval and another for result ranking. In this way, LinkedIn can more effectively match jobs with user queries.

In addition, LinkedIn developed a query engine that can generate personalized job recommendations for users. With the advancement of AI technology, LinkedIn is not the only company exploring the potential of LLMs in job search; companies like Google are also continuously innovating in this field.

In the past year, LinkedIn has launched several AI-based features, including an AI assistant to help recruiters. Deepak Agarwal, Chief AI Officer at LinkedIn, will discuss the company's AI strategy and successful case studies of its recruitment assistant at the upcoming VB Transform conference.

Key points:

🌟 New feature: LinkedIn launches AI-powered job search supporting natural language queries.

🔍 Accurate matching: The new system reduces mismatched job recommendations through improved search understanding.

💡 Technological innovation: LinkedIn uses model distillation to reduce computational costs while enhancing search efficiency.