 Live from Las Vegas, Nevada, it's theCUBE. Covering IBM World of Watson 2016. Brought to you by IBM. Now, here are your hosts. John Furrier and Dave Vellante. Welcome back, everyone. We are here live at the Mandalay Bay. This is SiliconANGLE Media's theCUBE. It's our flagship program. We go out to the events and extract the signal from the noise. I'm John Furrier, Dave Vellante. Go to ibmgo.com. If you're watching this, we have a lot of keynotes and theCUBE videos. And our next guest is Prashan Pooyan, who's this co-founder and CEO and chairman of Alphamotus Research, CUBE alum, back from our last cloud show at Interconnect. I said last year. I feel like last year, but past February, you were on. Welcome back. Thank you guys for having me again. Love the conversation we had last time. A classic entrepreneur story, right? But for IBM, it's a good one. You've been featured in a commercial. Now you're a celebrity. People want to get your autograph. Hardly. You walk in the front door. You pick up Lumix. You start playing around. You guys are data geeks. You're getting in partying with the data and you build a business. Now you're given the innovation talks. What's up? What's happening? I mean, you guys are really gone from zero to 60 miles per hour in such a short time. Where are you guys at today and what are you speaking about here at IBM? World of Watson. Well, the nature of the talk that I gave with Willy Tata was startup enterprise. So the idea is that businesses large and small, particularly in the financial services sector, need to start factoring in curiosity, creativity, imagination, experimentation into their processes because ultimately those are going to be the key factors that drive enterprise value. And startup enterprise because we started experimenting with the Bluemix APIs organically. We're able to train Watson on our domain expertise to understand our point of view on a particular topic so we can classify data that's coming in as a good or bad relevant to what we're trying to do. And now we're working with IBM to scale up the solutions and create building blocks for financial services firms that have the cognitive technologies on the back end and the financial services firms enabling them to find new and innovative ways to generate alpha. Take a minute to just describe to the folks who have not gone back and watched that video or don't want to go back and watch that video. They should go back and look at last year's video MinnerConnect. What you guys do, what the product is specifically, what's the innovation? Well, the problem in asset management is that many professional investment managers, many of whom are very, very smart people are having a lot of difficulty generating alpha these days. Alpha is beating the market on a risk adjusted basis and the reason that they're having difficulty beating the market on a risk adjusted basis, in our opinion, is because they're drowning in information. There are too many factors affecting the asset prices for the human mind to process. That cognitive dissonance is being compounded by innovators dilemma. What Clay Christensen is called an innovator dilemma. This idea that people are stuck in traditional business models and methodologies because they worked in the past and they're collecting fees but now they're starting to become a lot of, there's a lot of pushback from customers saying, well, I'm paying the fees, where are my returns? And so they're looking around saying, what's happening? You can't figure it out, but it used to be information would get disseminated asymmetrically. I would learn about something, I would tell you about it, you would tell your friend about it and that's how information will come into the market. But now, we all have our heads buried in our mobile devices. News comes out, we all know about it at the same time and we all push back our opinion. We run for the exit. And Mark, the individual cognitive biases of investors can compound on these collective manias and people are having trouble dealing with this. So when you deploy cognitive into your model into your system, you talked about classification. So do you have a predetermined taxonomy that then you train Watson on that or is that something that Watson informs you on and that taxonomy evolves? Can you talk about that? That's exactly right. We have built our own taxonomy using curated and balanced indications that we've been collecting from the floor of the New York Stock Exchange every day for the last five years. And that coupled with thousands of observations about supply and demand imbalances and stocks, coupled with observations about market volatility together with unstructured data coming in from thousands of sources, building an awareness cloud. We apply the domain expertise to the awareness cloud to create a point of view or a perspective and then fuel our building blocks, which are APIs that we then sell to enterprises. Talk about a data business. You are a data business. So you- Yeah, there's a lot of data. Presumably you start by understanding how data is, how you can monetize that data and then you've got to identify the sources of data and then you got to trust the data. Is that a sequence that you went through? Was that a linear process or? Well, yeah. I mean, so there are a lot of different monetization aspects to this. The data itself coming in can be duplicated and classified and in and of itself that becomes valuable asset that grows non-linearly over time because there's the element of making markets in human cognitively amplified human domain expertise where we're pulling in the APIs at one price from the utility provider, which in this case is IBM Watson, for example. We're adding our domain expertise and we're selling APIs out the other end at another price. There are a lot of ways to monetize. You monetize on the frequency of the updates. You monetize on the amount of data sources and the number of factors that you're looking at. So ultimately, speed of cognition with as many factors as possible is what everybody wants because that's what gives you the best estimates. And part of your secret sauce is not only identifying that supply and demand imbalance, but it's predicting which direction it's going to go. Is that correct or? Well, what it is that we're doing is we're taking the supply and demand imbalance indication, which is there's a distinctly human process for curating those imbalance indications based on customer interest. So the floor brokers will get a request. What's the early imbalance indication? Based on how many requests they get, that kind of drives their curation. The incentive model there is that they'll give out these imbalance indications because customers will then trade in the direction of the imbalance. So there's this feedback mechanism. What happens after that is, what happens after we take those imbalance indications is that we pair it with these unstructured data sources to measure deviations and sentiment. We combine those two together to create a volatility type of indicator, but instead of looking at volatility in price variants, we're looking at volatility in sentiment deviation. And the data source for that sentiment comes from where? Public data, news, text, tweets, images, videos and blogs. Most of the data is unstructured these days. And that body of unstructured data, the rate of information complexity continues to rise. You do it in a similar fashion than a human would. You're able to process far more and then analyze that and model that much more quickly. Right, what we're really making headway on is real-time cognitive analytics. And it's all about high-speed cognition. A lot of people think high-speed trading, high-frequency trading, it's really about high-speed processing of factors and data sources, turning the data into knowledge, applying a point of view to that data. And creating some sort of value. And now, one of our customers, if you're a sell-side enterprise, you might take our building blocks and create some sort of application that will add value to your customers to entice more trading. And on the buy side, you might pull these factors into your models to create better trading strategies. So ultimately, what we're trying to predict to go back to answer your original question is, how likely are prices to move in the direction of the supply-demand imbalance? When there's a spike in negative keyword velocity in contextually relevant information, sell-side imbalances are less likely to get paired off. So instead of just following the crowd, you can follow the crowd with conviction or go against the crowd. Basically. Essentially. Right, you can figure out what's real and what's just driven by sentiment. That relationship is very interesting when you start to build your awareness cloud and pulling more. I love this concept of an awareness cloud because you're bringing in this concept of cognition, but you mentioned real-time. In fact, I asked you last February, hey, what's going on with real-time? Because we were geeking out on the whole day of stuff because we have a data cloud too. We deal with all that, all that conversations. The observation space is the awareness opportunity. How have you changed or evolved Watson? You mentioned you're training the trainer or training the engine. Take us through what the innovation there is and how do you get to the real-time cognition? Sure, well, the first innovation is just bringing in a lot of data in real-time and applying these microservices in real-time. Things like the natural language classifier, the Alchemy API package, we have data sources such as weather now and price data, you have your structured variables, we have our imbalance indication data, the proprietary data. It has to all come in and work simultaneously. When we apply our domain expertise on top of that, that process is a little bit nuanced in the sense that you have to go back and retrain based on looking at your forecast and seeing how accurate those were. And if they're not accurate, you get feedback and you retrain. So there's a process that's very time-consuming, but that's ultimately your differentiator. So you do the homework up front, do the legwork train, set that up, and then it just runs. Yeah, I mean, once you give it, when you're pulling in information, you have to know if this information is good or bad, but if that information, to get the contextual relevance part, you might be analyzing spikes in negative keyword velocity and contextually relevant information, but the contextually relevant part is the hard part. That's the training, that's the domain, that's the point of view. And you have to be right, meaning you have to know what you're talking about when you're training these things. But it sounds like it's less of a data source challenge than it is an iteration of the model. The scaling is a data source type of challenge, but that's why we've partnered with IBM Global Business Solutions and some of this. And you guys are not a typical IBM customer. You kind of found your way into Watson. You got the great TV love, which is awesome, congratulations. Yeah, that helps. Tell that story, how you sort of stumbled upon Watson. Well, five years ago, I was on the floor of the New York Stock Exchange speaking with floor brokers, my friend Adam, talking to floor brokers how markets have changed, 15, 20 years they've been down there from their point of view. Technological advance, man, automation, all this stuff has come in. Started talking about, well, so what keeps, what makes you relevant at all down here? You know, there is value in the fact that guys down there who are still down there are staring at this one particular interesting event for the last like 15 or 20 years. You know, I mean, they have a domain expertise around the area of early market on clothes and balance indications. And I thought it was very fascinating because they actually, they had this information, so, but it wasn't organized, wasn't structured. We couldn't get it electronically, so we partnered with a few of the floor brokers, built a voice recognition-based data capture system. We started to collect and analyze the data, found that they were partially predictive of prices into the close of trading. But those forecasts, the forecasting ability of those imbalance indications, increased when there were spikes in what seemed to be chaotic negative news. You know what I mean? So, around that time, I saw Watson beat the Jeopardy champions on television, like everybody, and I thought, well, wouldn't that be amazing if we could analyze national language and scale to measure chaos and news and social media and predict markets or at least predict the direction or the likelihood with which prices are going to move in the direction of our imbalance? But, you know, it's inaccessible, right, to somebody who's watching it on. I mean, what is this? Probably Citigriple buy one or one of these big banks. Someone's already working on it. Yeah, the people with the largest, you know, pockets, you know, that's what these technologies are for. But I was playing around with a service called Alchemy API to parse natural language on small scale. Got an email one day from IBM Watson, from Alchemy, saying that IBM Watson had acquired them. Said, oh, this is interesting. That led to the discovery of BlueMix. Then I realized they migrated Watson to the cloud and they were offering it to developers of the service. And that was truly a paradigm shift because suddenly we could ask more sophisticated questions than ever before in prototype news solutions. You could dig in. Yeah, we could start to really experiment now, right? And that's what we did. We experimented with some stuff. CNPC started using, you know, one of our applications that was calling Insights for Twitter, analyzing, you know, a few hundred million tweets in real time for sentiment. You know, just a kind of a descriptive widget, you know, nothing really fancy going on there. But it was just interesting we could do it because we built the thing in like a day, maybe a couple hours. It's not a big deal. I mean, there's some, it was a little more involved, you know. But fast. But fast, relative to what I thought, right? And then started to explore around and, you know, I think the fact that we built something that was interesting in a new industry was appealing. We got on the radar. And then I started to realize basically that, you know, IBM could, you know, was actually working to cater to developers, you know, to the community to find use cases. They're not in the business of domain expertise, right? But they're creating this utility. Ultimately, the key differentiators in Watson, it's the human domain expertise and creativity with which you use these systems. So true augmented reality. And there's IP protection. And there's your mode. Yeah. Well, it's also the scale. It's also the augmented reality value that you're providing. Faster than the speed of thought, as they say, Bapachana once said. Prashant, thanks for coming on and sharing the congratulations on your company and continued growth. Appreciate it. Thanks for coming on theCUBE. Thank you very much. All right, we'll be back with more live coverage after this short break. You're watching theCUBE. I'm John Furrier with Dave Vellante. We'll be right back after this short break.