Do you remember our last article about Tinder Matching AI? Well, after a bit of research and analyzing the AI algorithm used there, we decided that it's not suitable for our purpose.

Let's pick up a concrete example:

Let's say Mark Cuban wants to find a CTO for his new Start-up Skell.io.

He is searching on Linkedin and he finds Teo Deleanu.

When we look through these two profiles we are sure that they match. But what about a robot? What is a robot supposed to do to match these two?

After analyzing the steps to create the Tinder AI, we can reproduce some of them but personalized to our own needs.

Let's start gathering data. On Skell.io, there are two types of people: Entrepreneurs and Developers and we will connect them through their Linkedin profiles. In the AI algorithm, we have to manually enter Teo and Mark's profiles. With that done, what's next?

We have to code the AI to search through their Linkedin profiles for differences. For example purposes, we can see on Teo's profile that he has a "Open to job opportunities" section.

But Mark's doesn't.

We were expecting those kinds of differences. Still, from Teo's "Open to job opportunities" section we can deduce that he's a Developer looking for work and we can save his profile as "Developer" but we still need an Entrepreneur.

Scrolling down on Mark's profile we find the About section.

Just what we needed, we can now save Mark's profile as "Entrepreneur". We can teach the AI to look in the About section and search for sentences like this, but there's a catch, people can enter whatever they want in here, so we have to keep looking.

Further down we find this "Articles & activity" section:

It might be a little useful, but not that much. Let's go deeper.

Finally, we hit the "Skills & Endorsements" section. Here we can clearly see that Mark is experienced in Start-ups and Entrepreneurship. We can tell the AI to look here as well to gather the necessary information.

Continue reading the story here.