Dmitriy Pavlov

So what we're able to do is kind of like validate and replace.

Yeah.

So to set the context, actually, I think it'll be helpful to.

So the original point of this of this technology was to look at survey methodologies being used by folks like the CDC Center for Disease Control.

And what the CDC would do is if they wanted to predict a region's susceptibility to disease like Miami's risk for heart disease.

What the CDC would do is they would go out and they would survey about a thousand people to get statistical confidence.

Takes a bunch of time, a bunch of money.

So what Dr. Johannes and his team did is they developed a new set of machine learning and essentially an evolution of natural language processing methods that are able to look at the linguistic structure and the syntax

of text.

And by simply looking at tweets, they were able to significantly outpredict the CDC for things like, we were able to predict things like, is the person depressed?

Is the person anxious about something?

Are they influenced by social factors like their family?

Or are they more influenced by social factors like their friends?

And what we realized is there's a really, really beautiful application for this technology in the consumer space.

We can look at what customers are telling businesses in customer support requests, in reviews, and we can actually understand

understand a whole new different way of understanding how severity works.

So if somebody, for instance, reports, hey, these three things are broken.

So for our first POC for this under-sink water filter that we did, a really neat thing came out that people were telling the company that, hey, installation is really critical to us.

The water flow is also really important.

And how much the water tastes is also super important.