Power up. From supercharging employee productivity and streamlining processes to inspiring innovation, Microsoft’s AI is designed to help you build the next big thing. No matter where you're starting, push what's possible and build your way with Azure's industry-leading AI. Check it out.
From vague notions about changing the world to hyper-specific employment situations, a new LinkedIn search feature aims to match your conversational thoughts with more tailored job results through AI.
The new search function will let job seekers type queries like, “I want to make cities more walkable and bike-friendly,” or, “I want to find remote entry-level jobs in video game marketing.” The aim is to surface roles that a classic keyword search might not have been able to suss out, according to the company.
LinkedIn VP of Engineering Wenjing Zhang told Tech Brew that the LLM-powered experience allows users to more freely express ambitions in plain language rather than contorting searches around certain buzzwords.
“The old search experience is…almost teaching you and molding you how to do a job search. We have all these keywords and filters. We’re almost limiting you,” Zhang said. “The bigger pivot here is, it is more empowering you now…We want you to express yourselves.”
Other new AI-powered features offer more detailed appraisals of whether a candidate’s profile is a good fit, and more transparency about how responsive the hiring company is.
The new features mark the latest way that the Microsoft-owned professional network is tapping generative AI to better connect its vast pools of candidates and hirers. In 2023, LinkedIn rolled out more conversational search tools for recruiters in 2023. Last summer, the company detailed how it was using behind-the-scenes AI agents to improve various features.
Opening up: The search tool has already begun to change how users interact with LinkedIn in early testing, according to Zhang. The team has noticed that searches tend to become more “vulnerable” as job seekers grow accustomed to the added functionality.
“They start to ask more expressive questions. Some are like, ‘I have this skill, I want to do [this],’ or, ‘This is my career aspiration, and I have this goal that I want,’” she said. “Some people will tell us, ‘I just want to make money.’”
Zhang said the new feature has even helped unlock new types of job postings that LinkedIn itself wasn’t previously aware of.
“There are jobs on the platform that we didn’t even know exist. There’s, like, taco wrapper jobs, burger wrapper jobs that are available, janitor jobs that are available on LinkedIn,” Zhang said. “Our system didn’t do a good enough job in surfacing those opportunities for those job seekers. So it unlocked us to go outside of the limited kind of tech-ish or white collar-ish type of jobs.”
Model mix: The search function is powered by a stack of different foundation models—both open-weights and closed—that handle different aspects of the process and are fine-tuned with LinkedIn’s own data, according to Zhang. That mix includes—but isn’t limited to—models from parent company Microsoft’s partnership with OpenAI, she said.
“We are definitely using the GPT model as part of the stack when it applies,” Zhang said. “If we can use the GPT model, we’ll definitely give it a try.”
As the feature rolls out, LinkedIn will be watching how relevant the jobs it surfaces are, how many people end up applying, and ultimately, how many candidates land jobs through the search, Zhang said.
“That is like a true north metric that we look at—whether you landed the job or not,” Zhang said. “But given that it’s more sparse data, and it takes a couple months for you to find the job and update your LinkedIn profile, we may not get that immediately.”