Even Computers Have A Hard Time Job Hunting

Poacht looks for your next job so you don’t have to. Which turns out to be a lot more work than expected.

Even Computers Have A Hard Time Job Hunting
[Young couple: conrado via Shutterstock]

Even if you aren’t digging through job boards for your next gig, there’s always that gnawing feeling there might be something better out there. And now there’s an app for those of you who are never satisfied with your current job: Poacht, an app and service which algorithmically tries to match a currently employed candidate’s LinkedIn profile with employers looking for the best talent.


Instead of using industry-standard keyword search, the company is trying to drastically improve how people look for jobs–or rather, wait for jobs to find them. “The algorithm looks at your skills, education, previous work history, employment level, location, and what we call a ‘predictive success’ value which uses a collaborative filtering model to weigh your chances of getting hired,” says Poacht cofounder and CTO Isaac Rothenbaum. “The outcome of this is a single value (with meta data) that highlights your strengths and weaknesses as they compare to the needs of a position.”

There are two main challenges the Poacht team has come across while crafting this service.

First, the skills people have listed don’t always line up with the job descriptions employers are looking for. Humans, with trade knowledge, understand that having Photoshop listed as a skill could fulfill a job description’s field for Adobe Creative Suite–computers don’t. It’s difficult to translate those types of discrepancies automatically.

To address the problem of translating skill between resume and job description, Poacht is actually doing it manually. The company hasn’t found a reliable way to automate the task yet, so until then it’s relied on the interns to help do it by hand.

Second, developing an algorithm for use with primarily static data isn’t easy. The discovery engine for a service like Spotify or Pandora, for example, gets plenty of data to use from song choices, skips, and constant user interaction. The problem for Poacht is that once a user has filled out their LinkedIn profile nothing much happens.

The company is now looking into algorithms for handling sparse data. It’s also working with a PhD candidate at Johns Hopkins who’s researching applying machine learning techniques to big data in health care to address the problem. But in the end, there may just need to be more interaction created with users.


“We do have some ideas in mind, maybe tying in gamification elements to encourage user buy-in, but they are all very early stage,” Rothenbaum says.

Poacht has already updated its algorithm since going live about a week ago. And version three of the app is already coming into focus.

About the author

Tyler Hayes is a Southern California native, early technology adopter, and music enthusiast.