MatchAI is designed to help job seekers and employers find each other. It uses API integrations to help employers put their best foot forward, and artificial intelligence to help users discover new and exciting job opportunities.

APIs used

  • ZipRecruiter job postings
  • Scikit-learn (machine learning library in Python)
  • Yelp (Search API)
  • Google Maps API

The problem -- 'Long Tail' job postings

The massive amount of job posting data out there leads to the 'long tail' problem -- there are tons of smaller companies and postings that don't get as much attention as large and established players, but there are a lot more of them to choose from.

Any long tail problem can benefit greatly from machine learning to help rank what a user would find interesting among the massive amounts of choices in front of them.

MatchAI is inspired by some of those apps, and helps both sides by supercharging job postings, as well as surfacing only the best suited jobs for each job seeker.

How it works

Job postings get super-charged with APIs:

MatchAI uses a couple API integrations to make job postings more interesting.

  • First, it uses the Google Maps API along with the user's location (from their mobile phone) to calculate their commute time to the job. This is an important piece of consideration for job seekers that was impossible to do before GPS; everyone that uses the app of course gets a commute time specific to them.
  • It also uses Yelp's Search API to surface restaurants and cafes nearby the job location, to better sell the job to potential employees. Interested applicants and play around with the included map on each posting to check out the nearby amenities and area of the job.

AI helps to rank jobs to show, as well as WHAT to show from each job:

MatchAI uses AI in two ways to make the job search process more engaging and successful:

  • It uses a machine learning ranking algorithm to rank which jobs to show users based on the users swipes in the app. As the job seeker sifts through more and more postings, the model learns what features of each posting -- the commute time, salary, etc -- are important to that user, so it re-ranks what to show them
  • It uses similar feature selection algorithms to find what pieces of each posting the user likes. For example, do they click on the info button to dig into the job? How long do they spend playing with the map vs. scrolling past it? Do they ever make it to the 'culture' section of each posting? As the model learns this about each user, it can vary what to show that user to best sell them on the position.

Target User / Business Model

Though this could be a stand-alone app or product, it would benefit greatly from job posting data -- something that a company like ZipRecruiter has in abundance.

An example use could be a Premium Posting Upgrade when an employer posts a job posting. In return for paying the premium, they know that ZipRecruiter will 'spice up' that job posting with pictures, location data and the such to help sell it better. Tinder did something similar with 'Smart Photos', which lets users allow Tinder to pick the best photos for their profile based on actual user swipes.

Though companies like Jobr and Switch use a 'Tinder' like interface, they don't have the posting data to capitalize on the AI aspect of what Match AI does -- that requires a LOT of data, and a larger company in the job posting space would be best positioned to take advantage of this technology.

Future Plans

The real magic here would be to use more sophisticated personalized data for each user to structure job postings differently based on who's viewing them. Knowing exactly what a user appreciates in a job or is looking for, will help a company like ZipRecruiter better surface the best jobs and do so in a scalable, automated way.

Share this project: