Inspiration
Job searching often involves sifting through dozens of different sites, each with their own confusing filters and settings. LinkedIn alone has more than a dozen different filtering options, making it hard to find what you're looking for. We wanted to make this tiresome process easier and more intuitive—so we've created a browser extension that allows you to simply tell a chatbot what you're looking for, and the filters will be set for you. We would like to combine LinkedIn's search function with an AI chatbot. This will allow job applicants to conserve time and energy, thereby being able to focus on real career prospects instead of fiddling with complicated search options.
What it does
Slide-n-Seek is a Chrome extension that uses OpenAI's API to help users discover the most relevant job listings on LinkedIn. It uses a technique known as RAG (Retrieval-Augmented Generation). First, the AI interprets the user’s job preferences or description. It then takes that initial output to refine the search criteria, automatically applying filters and directing users to more focused job results based on their original query.
How we built it
We decided that to improve a current application one of the most intuitive ways to do so was to create an extension for the user as to not disrupt the apps intended design. This way we can rely on the apps already planned out UI/UX and instead just inject our improvement in a way that does not take the user out of the intended experience. Making the structure of the extension was relatively easy as it only requires 3 base files. We decided that we could help the user even more by letting them describe their ideal job using natural language that way the user can be very comfortable with the functionality. We have an AI Model (ChatGPT) analyze the user's input and generate a JSON filter encapsulating the user's core desires in a job. This way we can the generate a URL to a filtered job search page on linked in for the user to search their most relevant jobs just from their description. This last processes uses the RAG approach to prompt engineering as we move outputs from a model as an input to another model to refine our end results.
Challenges we ran into
Initially, we agreed to modify LinkedIn directly but ran into multiple issues with LinkedIn and privacy. It turns out we were doing a form of web scraping that linked in explicit attempts to prevent, so we moved on to alternative options. Another major problem was injecting Javascript code into the website to modify it. Fortunately, LinkedIn allows our extension to redirect users to the location we want the user to go to without altering the website's source code.
Accomplishments that we're proud of
We are proud of being able to create a workable and user friendly extensive that is styled somewhat similar to LinkedIn's core design as to not take the user out of the experience. We are proud of the research we were able to do and work in a fast paced environment and adapt quickly to challenges and overcome obstacles. We are very proud of the end result as it shows that we were able to make a good extension despite challenges that could have easily dissuaded, our motivation to make something that we would genuinely use in our everyday lives.
What we learned
We learned how Chrome extensions work and how to alter web pages with content scripts. We also used OpenAI's API to get real-time responses from a chatbot. We learned we need to carefully deal with different ways of running code and to keep filter elements aligned in LinkedIn's changing web page structure. The project showed the importance of good user experience design. This includes making the chatbot conversation simple while dealing with many LinkedIn filters. It also showed how smooth AI integration can make a difficult job search much easier.
What's next for Slide-n-Seek
Going forward, our top priorities include broadening Slide-n-Seek’s reach to more job boards beyond LinkedIn, expanding coverage of advanced filters such as company size or benefits. We also plan to introduce a more robust conversational flow so users can refine their criteria with follow-up questions. Finally, we want to add persistence and syncing features (e.g., saving “dream job profiles”) and refine the UI with animations or onboarding prompts.
Log in or sign up for Devpost to join the conversation.