 It's theCUBE, covering HPE Big Data Conference 2016. Now, here are your hosts, Dave Vellante and Paul Gillan. It's back to Boston everybody. Mutisha Dunda is here, he's the founder and CEO and he's joined by Michael Bishop, who's the CTO of Logitbot. We're going to talk about seeking alpha, gentlemen, welcome to theCUBE. So, tell us about Logitbot. Why did you, you know, found the company, what's it all about? So, the short story behind it is that, I've sort of been in financial services for a really long time, did my tour of duty across Wall Street and almost 10 years ago now, ended up at a very large market-making firm in the Philadelphia area and was the chief of staff of their proprietary trading business. What that business was going through at the time was this evolution where people are embracing high-frequency trading technology, sort of replacing human traders with automated software. My company at the time was very resistant to that change and sort of valued the insight that a market-maker or human could provide during times of volatility. A lot of our competitors sort of threw away a lot of their traders and when the 2008 sort of crisis hit, we were one of the few people willing to be buyers and sellers of last resort in the financial markets and it was the best year the firm ever had during that period of turmoil and the reason was human beings could make rational decisions. If somebody came and tried to sell you a blue chip stock at 80% off of its value, someone could sit there and say, I don't think this makes sense, I'm a willing buyer. So, that sort of allowed us to basically have a trader not worry about the weeds and focus on the big risky trades and have a machine interact with kind of the low order flow, kind of day-to-day order flow and that is what sort of got me into the space. I left the firm and I joined Bloomberg and I was the head of strategy and business development for five years and the company is amazing and they've built a business around providing people information. What's happened in the information business is that people have gone away from static data to analytical data and so now what we're doing at Logitbot is combining the ability of machines to sort of take the grunt work out of your day-to-day job and allow you to focus on high level things and then also provide you with very useful insights using the power of modern computing technology. So we're combining a helper with intuition and an ability to deliver analytical capabilities. So the money business and the tech business have collided in a big way and you're taking advantage of that by providing a platform for insight. Is that right? Okay, and you are, are you an alpha geek or are you not an alpha geek? I guess I am. And I say that with all respect. I wish I were an alpha geek. I have a very sort of scientific background as an electrical engineer, data masters in financial engineering. So I approach financial markets with sort of that systematic or analytical perspective. And I think in today's world, it's sort of the one place you can go to in periods where we're in an environment where people have never seen this before. Interest rates are negative around the world. Like there's no precedent to a lot of things that are going on. So you can actually apply science to try and understand how markets will evolve, given all that's going on, using a process. And you, Michael, are a platform builder, I presume, and a technologist and you create products. Is that right, or as well as potentially an alpha geek? Yes, and yeah. Okay, so tell us about your platform. Yeah, so. You build it, you build it. So our platform ingests an enormous amount of data that it has to reason about and distill that down into insights and predictions for our end users. And that, you know, that spark and using BigQuery, Google and TensorFlow and HP have it on demand to do some entity extraction on the news that we process. So lots and lots of volume, lots and lots of data going through. It goes to the meat grinder. Do you, so, okay, so yeah. And it's a lot of secret. Everything you do is a trade secret. Yeah, there's a lot of secret sauce in there. There's a lot of secret sauce there. At a high level, just sort of what to, the way I would explain it is that a lot of information is generally publicly available. You know, large companies are forced to disclose material information. They file regularly with the SEC. And if there's anything material, they actually have to have a press release, you know. So information is generally available, but it's in a format that a lot of people can't make use of given the volume of information. So if we can apply kind of artificial intelligence technologies to extract information, summarize it, find value that is hidden in text, you know, things of that nature, we can kind of give our clients an informational advantage. Again, in the financial services world, that is all we trade. The currency and the business is information. So allowing people to access information easier, faster, better, have a much more clear picture of what's going on in broad markets is obviously extremely useful. So it's hidden information that you, it's there, but it's just not apparent is it text mining that you sell at? Yeah, that is actually a good core strength of ours. Another one is this notion or technology that we use, which is a connected graph of the world's financial companies or actually the world's publicly traded companies and all the important people that are associated with those companies. So as an example, again, you could have a technology company that has a supplier in Taiwan that is delivering a component that goes into a phone as an example. And if that supplier in Taiwan starts indicating to its Taiwanese investment community or the regulators there that it has an inventory buildup, right? You could have a piece of technology that could read that and make an inference about this US-based customer on theirs, right? So in the case of a phone, they could say, oh, these phones might, the sales of these phones might be weak because we were expecting the Taiwanese supplier to be selling the components like hotcakes and it's not the case. So in a previous life, I did that for the disk drive industry. I mean, I had the whole supply chain figured out and it took me years and years and years to build. And it took me a lot of phone calls, a lot of dinners, a lot of drunk nights getting information and I wasn't so drunk, it turned out, but they were. And but it would have been impossible to scale, right? So you figured out a way to automate this whole. You can obviously use the power of even technologies like Haven that do text extraction, entity recognition from raw text. So you could have a computer read text and then extract, this entity is an entity and it has these products and these products are supplied by these other entities and kind of build that graph without having a human being having to collect the information manually. You have a potentially infinite amount of data. Yes, that is right. That could be relevant to what you do. So how do you apply big data principles to actually narrow that down and find the data that matters? There's a lot of filtering. As you might imagine, there's a lot of junk in what you bring in. So there's an enormous amount of effort that goes into filtering out things. So for instance, with news, we're ingesting 70,000 new sources which have been curated and ranked by hand for importance and reliability and that's constantly updated and checked. And there are automated ways in which we filter things out and throw things out. The other thing I'll just add is that the other way we do it is we just try a lot of things because again, with a computer, you can almost run 50,000 things and see what actually is meaningful. So that is also another really easy way of solving a problem is you have the ability not to just try one thing. You could simultaneously try 50, 100,000 things and then understand what's really going on based on those 100,000 experiments that you just did. In the Haven portfolio, you use a little bit of Hadoop, a lot of autonomy, a little bit of Vertica. We are big users of their APIs. So what they have there is almost like little building blocks for machine learning and nifty things with text extraction. So we like that because you can actually design your own, I don't know, software ingestion platform. It allows us building blocks that we can quickly run experiments on in the new combinations product that they've announced. We had a preview, we were able to work with a preview of that and we think that's going to be even more of a force multiplier for us and it lessens the cost in time for us to run experiments and it also pushes, it pushes the complexity or eliminates a lot of the complexity of running them down so that we're able to push us down to a data scientist or someone who wants to run a machine learning experiment and they can do that with little to no coding involved so that that's a dramatic improvement over having to spend a few weeks coding something out. We've got this massive API economy that's developed and I would think that a company like yours would thrive off of that. I mean do you plug into a lot of APIs to assemble the data that you process? Absolutely and we also are providing APIs of our own. So yes, just to answer your question, we definitely believe in the API world. We are able to leverage technologies that would have taken a firm like ours decades to develop or even scale. So through the power of APIs, we're able to kind of quickly turn things on and make the calls of retrieving information we want very seamlessly. We're at a time where through API delivery, it's kind of democratized a lot of these very expensive to purchase on-premise technologies, things that were just out of reach for a small company even just a few years ago. Can you give an example? I mean what kind of data are you able to plug into now through an API? So I would say for instance, entity named entity extraction at the level of accuracy that we get with Haven on Demand particularly is something that even with our very, we have a given that we have a vast database of entities, you would think it'd be very easy for us to do the same thing. It's a difficult problem to tackle and HP has helped us considerably there. Add to what Michael's saying, so a good example might be, so we're doing this across 120 million chains. So that's the first piece of complexity. The second one is if you're trying to resolve Apple Inc the company and you just read Apple versus the fruit versus New York City which is often called the Big Apple, it's that really hard problem for a computer to solve. So in doing that at scale again is not trivial. And we're doing this across, we're streaming thousands or hundreds of thousands of news articles every hour and analyzing that data and then making these categorizations of I read this piece of information, it was about this entity, a company and it was taking an action on this other entity, a person or a company and it was related, so you start connecting the dots, it's not easy. And those steps are critical in reasoning about the data. It's not something that we necessarily need to invent in-house that frees us up to work on core algorithms to interpret the data instead of spending all the time trying to disambiguate Apple from the Big Apple for instance. So it makes it, that's a, our day to day is dramatically different. The problems that we solve are dramatically different. We could spend two years working on named entity recognition and getting that very accurate or we could plug into an API and work on algorithms, financial algorithms that do actually, that matter to the company. Absolutely. And you sell a subscription? So we are, because it's an institutional product, the data is made available via API. So it, again, it depends on kind of the application that the information is going into and whether or not the information is going to a human eyeball or to another system or machine. Okay. So it's a subscription based business but it depends again on kind of the entity. Yes, that was my next question. It's how is it consumed and you're saying it's consumed differently by machines than it is by humans. Absolutely. Okay. So we have this, for human consumption we actually have a really nice AI and he has a name, he's called Rob Otto for Roboto but you know his name is Rob. And... Mr. Roboto. Yeah. And he is a fairly intelligent financial analyst. So you can ask him questions about markets, stocks, trade ideas and he will try to answer you back in a human. So you can have a conversation with Rob and what he's doing in the background he's actually doing a lot of financial analysis and coming back to you with a response that a human being would expect versus a machine that would want structured data or just analytical information. And you built this? Yes. This is, you guys built, that's really your secret a big part of your secret sauce as well. Yeah. And so it's in a way competing with Watson. Yes, is that? Absolutely. And not necessarily, it's made by everybody. The secret sauce. See absolutely as I can ostensibly anyway talk to Watson. Yeah, yeah, yeah. Response to not necessarily is Watson doesn't can't do it. Yeah, Watson is really good in healthcare and a couple of other industries. I don't know if he's been trained up in financial services. And again, we are very full, that's all we know. I mean, that's sort of our background. And in machine learning, one of the traps that a lot of people fall into is you think you can go get the best model or neural network and it'll give you the best answer. And what turns out to be the case is that having that kind of context, industry experience, intuition around what's important is really with the secret sauce. So we've spent over 35 years in the business. We kind of know what financial people care about. And if you asked our machine a question and you asked another piece of amazing AI a similar question, you might be surprised because we tend to be smart. You think of sentiment and you pose to a sentiment end point that is not trained in financial terminology and you give it an article that someone has issued shares. A company has issued shares or bought them back. And what is, is it good or is it bad? And, you know, if you ask a generic one, it's happy or neutral or you know. They don't understand. There's no context there. It has no idea what that means. And so our, that's something that we have to do in-house because a sentiment, a sentiment end point for instance is provided as just a, it's a generic sentiment thing. It's good perhaps if you're looking for tweets about Jennifer Aniston or something. Yeah, right. So in that example, a very simplified example about happy. Yeah. Issues versus dilutive. Exactly. Impact, you know, long-term, near-term. Interest rates went up. And, you know, typical machine learning thing will tell you, oh, this is good. It's up there. Why is the market going down? Yeah, and you're like, no, it doesn't make sense. So, yeah, that's kind of, so having that specific kind of industry know-how is really important and you have to instill it into the machine because it doesn't know these things. Now, how do you price? Now, I know it's got to be really, really expensive because of the value, but what's your pricing model? Is it by machine, by user? I think it's by analytical category is the best way to think about it. So in the unstructured data world, we have a product that basically monitors news for you, connects the dots, tells you what's important and what's going on and how that information is getting added to our knowledge graph. So that's kind of one category of data. Another one is we have predictive models. So you can actually say, given all these things that aren't conventionally part of an econometric model, say, like sentiment, news, political risk, what central bank is saying, like how do you capture all of that into a model and then allow me to predict stock returns or the price of a specific class of securities in the future? So those predictive models, again, are like a different solution or a different product. So we price by that. Okay, more on the company, like how long you've been around, funding? So yeah, we are young. We've been around for about a year. We are self-funding right now, but we are in the process of actually going through our first round of outside money. So we're talking to people now who have shown a lot of interest. New York, VCs mostly. Yeah, mostly in New York, because I think they understand what we're doing and what we're all about. Yeah, and I think there is a pretty robust community around the FinTech, of investors around the FinTech space in New York. How many people are you today? We're four of us today. Awesome. Yeah. That's fantastic. And you've got, you're live in the marketplace. You've got real customers. Yeah, we have customers that are testing and using all products right now. So you guys self-funded, but you're generating maybe not positive cash a little bit cash, right? Yeah, a little, yeah. And it's again, like with what we're doing and what we fund and even the reason why we invented Rob was because it's still a new field for many people and financial services tend to be conservative. So the sales cycle to a large enterprise is obviously challenging and you have to kind of explain what you do to many layers of decision makers within an organization. So if we can rather than explain show value is sort of the tactic we've taken. So we're very happy to partner with people and kind of prove out that what we're doing makes sense for them for we try to commercialize product. With all the experience that you have though on some of the companies that you worked with, worked for, you would have a foot in the door at companies. Yeah, it helps. It does help. Customers. The, I wonder about starting a company these days and you know, you're four people and you're building a company. And it seems like five years ago that would have been almost unthinkable. Are there any infrastructure investments involved in what you do or is everything in the cloud? Michael can speak to this. Well, having the cloud, having cloud providers is really a game changer. But we still, it may surprise some people, we still do use bare metal for 30% of the infrastructure. And why is that? Why? Performance. Raw performance. Or there are only so many ways to get a terabyte of RAM. And it's right, yeah. All the caching in the world. GPUs, you know, power eights or that, you know, things, there are different types of architectures that you just can't get virtualized at the moment. And so you have to kind of rely on big machines, big metal machines. Is there something that you get out of using vertical that you couldn't get through any other source? We're not. We're not actually vertical customers, but we are HP AVEN on-demand customers. Which is a sort of blend of. Vertica and Autonomous. Yes. Yes. And Hadoop. And you're disruptive, you're disruptive, you're disrupting presumably Bloomberg. We're definitely trying with, I don't know. I mean, I know you probably don't want to say that. Spec and deal, yeah. You're so huge and a giant, but I mean, we're definitely trying to change how people go about doing their day-to-day jobs. That's kind of our focus, yeah, so I think. Pick up the phone, they work in spreadsheets and they look at things one at a time. So we're trying to change a lot of that where we can automate what you do on a day-to-day basis. What we're trying to do, it's very difficult to go deep at scale. So if you're trying to model a high-performing research analyst, even a high-performing research analyst cannot cover more than a few stocks. Maybe a sector, maybe. And they're only able to revisit that once a quarter, maybe. Or when Apple gets a giant fine or something. We're able to do that at scale for everything all the time, 24 hours a day. They're living in 10Ks and 8Ks and docs and reading the notes and it's just torture. And we burn through those and ingest them, rip out. No, no, no, no. Yeah, exactly, filter, filter, filter. And then we find just what we need to zoom in on and then that piece of information is either acted on immediately or it's added to the graph for potential later inclusion. Because it could be part of a very long trend of events where we need to go back and follow the breadcrumbs to an original event. And so that's added into the event graph. Awesome, we got to go. Yeah. Great story. Thank you so much for coming out and best of luck to you guys. Appreciate it. With the raise and hopefully see you here next year with a continued story. Yeah, thank you so much. Thank you guys. All right, pleasure. Thank you very much. Keep it right there, everybody. Paul and I, we'll be back. We're going to wrap up day one here at HPE's Big Data Conference. This is theCUBE, right back.