 Okay, we're back, thanks for watching everybody. This is theCUBE, Silicon Angles coverage of IOD. I'm Dave Vellante, and we're here at the Mandalay Bay in Las Vegas, covering IOD, wall-to-wall coverage, as usual. You can tweet us, I'm at Dave Vellante. My co-host today is Jeffrey T. Kelly, K-E-L-L-Y. So please tweet us and let us know what you think if you have any questions. So I'd like to introduce my co-host now. Who's going to introduce the next guest? Jeff? Welcome, everybody. Just quick correction, Jeffrey F. Kelly. Well, Jeffrey F. No, T, I don't want to mess that up. So Jeffrey F. Kelly, K-E-L-L-Y. Right, important for the Twitter handle. And then you've been, welcome everybody. Thanks for watching today. First guest, very interesting company, Bill Hartman, president of Terra Echoes. Welcome to theCUBE. First time. Appreciate Jeff, thanks. On theCUBE and we appreciate you coming on. Yeah, thanks, Bill, for coming on. Why don't you set it up? We were talking off camera, and somebody made the comment, people around here care about data, you made the comment about your customers care about getting value out of data, getting insights out of data. So tell us what Terra Echoes does, and then we'll get into it. Okay, well, the short story is, and Jeff made a little distinction about real time just a minute ago. What we do is true real time, live. Our applications are all about where a customer has to determine in the moment what's going on and then be able to take action on it. So it isn't, the speed at which we do processing is just means to an end. And so our applications, we started with a problem that was incredibly time critical based on multiple sensors, and we didn't have a way to solve that problem until about four or five years ago when we came across IBM InfoSphere Streams. So it's a streaming technology, there are other streaming technologies out there, but what's unique about it is it was designed and it's exceptional for not just streaming data, but massive amounts of big unstructured data. So what we're able to do in that is take, just as a real quick example, cyber and a video and let's say social media, but let's say there was some application where those three particular things are available but haven't been able to be brought together before. What we can do in that, and what we've already demonstrated to our customers is on the fly, continuously taking those three different sensors, if you will, analyzing them and then acting on that information in a continuous moment. So what we're not doing is we're not persisting to a database, what we're doing is the data never stops. We actually are analyzing it and acting on it on the fly. In real time. In real time, true real time. So take us through some examples of how customers are using this. Well, we're a startup, so we're early. Most of our focus has been on the, what's called the intelligence community, the IC, and the Department of Defense where, if one was to go Google their sensor problems, they're literally drowning in information and they can't get to the information fast enough. It's just all sort of dropping off the table, if you will, given their traditional ways to analyze it. And so what they need is a way to achieve time critical insight into the information and then be able to act on that. Not everything, but there are certainly times whether it be in cyber warfare or whether it be in protection of critical assets or something tactical in the field. Picture a cobble or someplace like that where you'd want to be able to know if you're the guy on the ground, CL Team Six, let's just make this fun. And you were doing something where you wanted to know you've got the best available information and the insight from that information before you bang the door down. Okay, so you essentially have a software platform, a software application, how would you describe it? What we do is we're building on InfoSphere Streams. So InfoSphere Streams is a fabric that allows us to then build on top. We develop our own environment on top of InfoSphere Streams and then we develop our analytics, either proprietary analytics or we take analytics that are already out there but that are developed for the serial world and then we adapt them into our environment and tailor those to the applications. And what's the business model? Is it, you sell it as a service? Well, right now, eventually what we want to do is get to selling it as a service. I mean, there's some technical reasons why today where we could sell it as a service but it basically, it's off a cluster as opposed to a cloud-based kind of a service but as a technology, both the streaming technology evolves and our ability to adapt our analytics to it, we would see, let's say maybe three to five years from now, analytics as a service being what we're doing. So if I understood correctly, you're not storing the data. We don't store it. Now, to your point about working with integrators, certainly, I want to put in a plug for Netiza. I can see why Rob Thomas thought was impressed with the team. We've done some early work with Netiza and they're incredible people. We love their team and we love their technology but so for example, we could be working with a Netiza device where because we don't stop the data but we might want to do some deeper analytics to inform us, let's say five minutes from now, about something that we need to look at. We can modify our models on the fly based on Netiza or a green plumber, some other device informing us. So there's a go-between between the two of us which would be similar to, we could talk to Vivizimo, we could talk to Palantir or traditional database kinds of resources. So we're not only taking live data but we're taking all data as if it was live. Interesting, so the way I kind of envision the scenario is as the data comes in, kind of that streaming component is kind of the entry way and that data kind of flows into another, either a database like a Netiza or maybe a Hadoop or whatever it might be for either deep analytics or more SQL type analytics. And then as you mentioned, you can feed that back into the streaming system to take advantage of what you've learned over the long term and incorporate that with kind of the in-the-moment analytics. Right, and I think of this very crudely as the canary in the coal mine. I mean, there are clearly applications today where you need to know as fast as you possibly can. But we've also talked to people that are in industrial control systems. So picture a power turbine, a major industrial-sized power turbine where if you could manage all the different sensors that are on that more precisely than you can now, truly in millisecond time based on looking at 400 sensors, all in the moment to say, one sensor would have said speed up, but really what you need to do is slow down just to make it real simple. If they could shave maybe a half a percent to a percent more in the efficiency of that turbine, the investment pays for itself almost immediately. So we see a number of different applications but it's such a different way about thinking how to actually use data. To me, it's actually more obvious what data is worth if you're using it real time. But what we're used to, what the world's used to is putting it into a database and then somebody has to ask the question and then was that the right question? What we're doing is really asking the question, we're never not asking the question. We're asking it all the time and then adjusting the question on the fly as well. So yeah, and a lot of that was due to technical limitations that certainly weren't just wasn't possible until recently. But I want to talk a little bit about kind of the sensors themselves. We heard about 10 years ago RFID kind of hit the scene, got a lot of press. It was kind of going to be the new thing. And then we kind of died down and now we're kind of back to the internet of things and things being censored. Where are we, you've got your feet on the ground in this space in terms of connecting all our devices. I mean, where are we in terms of is the average industrial component censored at this point? Is the average consumer products censored at this point? Where are we in terms of actually getting those physical sensors there so that we can actually grab data from them? Yeah, well realistically, there's a continuum. I mean, there's people on the cutting edge and I was at a breakfast here with IBM the other day and one of the IBM champions who does a lot of consulting, the Fortune 500 company said, she has Fortune 500 companies that are still doing a lot with paper. So even in those kinds of companies, they're not the internet of things yet. But in general, whatever, how many, eight billion cell phones or something like that or smart phones and where all that's going. In our early applications, where we're going is where the sensors are already there, clearly already there. It might be in the military, it might be in the Department of Defense, it might, you know, satellite imagery, full motion video, you know, what the military and the commercial world has really come to like in the last 10 years. I mean, you know, we're on YouTube is video, more and more and more and more videos. So the military is using it a lot, businesses are using it more and more. I was talking to you off air about a company, a commercial company that's using facial recognition kinds of technology to recognize their shelf space and what's on the shelf is placed properly or not. So to us that facial recognition input is a sensor. You're global positioning us, the three being right here and that we're static for a certain amount of time and being able to monitor what's going out on the air or whatever would be a couple of different sensors. We also monitor cyber. So for example, this isn't speaking to the broad market but you could picture something like a Stuxnet in a broader sense where someone is trying to attack a facility, but nowadays it isn't a single vector kind of an attack, it would be cyber from multiple different angles if you will and physical from multiple different angles and all of those things are likely in an industrial setting to have some kinds of sensors on them. It might be a fence, it might be video systems, it might be how your system is operating and you take all that and you put it together, it's going to tell you something that you wouldn't otherwise know and in some applications that's critical to know and some it isn't. Whether we could help Nordstrom or Target help somebody buy something more quickly, I think we could, but whether that's worth all, the ROI probably isn't there yet. So your focus today is really on solving those hard problems that are related to potentially national security and other factors. Talk about what your clients and how you envision your clients accommodating this new capability because you started off this segment saying your customers have all this data that they're ingesting and they have no way to really process it. So how are they changing their processes to be able to manage all this or actually get outcomes, describe the outcomes that they're getting and how are they changing their processes in order to be able to exploit the capability. And I wouldn't want to overstate how we're driving customers to change processes just yet. We really, how do you envision it? We pivoted this, yes. So how we envision it is our customers, particularly in the Department of Defense, the intelligence community already have a very serially oriented workflow and that workflow moves from sensor input and it might be multiple sensors. They call it intelligence surveillance and reconnaissance but the analogy to what happens in the commercial world I think is exactly the same in BI where what they're doing is they're serially processing based on some sort of data-driven, whether it's a machine-driven or person-driven decision to sort of do 80, 20, what's important, what's not important. And that's what's breaking down right now is there's so much data coming through they're not able to figure out, gee what did I just miss? Someone yesterday here at the show was talking to me about a subway overseas where someone recently committed suicide in the subway and the subway system like New York, London, Singapore, where else has somebody or four or five guys in a room looking at 30 or 40 different monitors trying to keep up with what's going on. That just doesn't work and so now there's, he was telling me a backlash in this particular city where they've got this video over and over and over again of somebody committing suicide despite all this technology not being able to get the information about what was about to happen and perhaps say this man's been loitering there or he's doing something unusual, why don't we have someone go over and take his arm or ask him if he needs some help or something like that. Not I don't know if that's exactly the right application but there are workflow, my point is there are workflows already out there and what we picture is plugging into those workflows not telling them blow your workflow up. Really, again in the Natisa example of the Vivizamo plugging into an approach that's already there but making that approach much better. So that's a huge advantage obviously for your business because otherwise you're going to delay the adoption. Eventually this is going to change the world but we got to take baby steps. Talk about the company bill. You mentioned your startup, can you talk about where you're at, funding, things of that nature? Sure, we raised, thank you. We raised a seed round last year led by Flywheel Ventures and True Ventures. True Ventures and Flywheel both very excited about big data and saw the opportunity with what we were doing with this stream's fabric and the analytics to really be the old thing about catching the wave. So we think we're right there and really excited about it and we'll probably be going out in about 12 months for our A round. For an A round. For an A round. So we were seed stage. Seed stage. Can you talk about how much you raised? A couple million. Okay, so sizable seed round. Yeah, seed round and we got traction with several customers. Our initial customer we delivered a system to in January. Pilot system and we have a follow on with them that's a significant development for the next year. And then we're also talking to two or three other government customers and a couple of commercial applications. So we hope to see several more orders in the next two or three months that would help lay the groundwork for the A round. So what's the 100 day plan? What's the big milestones to get ready for that A round? What are you really focused on? I'm hopping on a plane tonight to go to Washington DC and it's not a 100 day plan. It's a two or three week plan. It's a 100 hour plan. Yeah, now we've got several things teed up where just like in the commercial world, there's no problem getting meetings with people to talk about big data. So we're talking to all the name brand intelligence community agencies. And again, we're not selling to all of them but we're talking to all of them because they all have their verticals. I mean, we probably think of the U.S. government as this monolith but what the FBI needs versus what the CIA needs versus what the National Reconnaissance needs are all completely different. They're all similar workflows but different. And so with one, what we're focused on is live, what we're doing with them is live feature extraction from full motion video on the fly. So picture of the video is always coming with multiple channels, let's say 32 channels of full motion video and what we're going to develop for them and the analytics is to be able to pull out particular features that they want to see as the video is flowing by. Certainly we see down the road lots of commercial applications for that but right now this is a workflow for them that's just bottlenecked. Is there, we had Jeff Jonas on yesterday he was talking about geospatial. Is there a play for that in what you guys are doing? Yeah, we actually, we under emphasize that the geospatial part ties into all we do. Our company actually has a very, very strong GIS background and we see the play, not the play but the need for geospatial because you gotta, you can't just analyze the information, you gotta act on the information. If you don't know where the information is you're both in cyber, there's actually kind of some neat things about cybergeo where things are in cyberspace but certainly in the world we all get right here what underlies almost everything we do is the geocomponent which also is what we're doing in real time. Live real, true real time. How much discussion, engineering thought do you guys give to privacy in that context given that geo is so fundamental to what you do? We haven't given it much thought yet particularly because we're plugging in the workflows that are already there. I mean there are bigger issues like what's going on in the Senate and the House of Cyber and policy issues and all that. We're a startup, we're focused on the initial applications and not getting, I don't want to say we're self-insuring but I think a lot of that's going to work out and particularly for the applications we're in now privacy isn't the issue. It's really, if you will, proprietary sensors that the customer owns. Well I think that's certainly going to be an issue going down the line. I think another issue when it comes to the type of streaming real time analytics we're talking about is it's one thing to for the system to tell you okay something has occurred and another thing to be able to respond. Oh sure. To actually take that action. A, know what it is and B, being able to actually execute. Sure. So you're still early days and it sounds like the clients you're working with have a very specific use case. But how do you envision helping clients in the future actually when you start maybe perhaps working with retail customers or others who this is kind of a new type of thought process for them. How do you envision kind of helping them not just transform their data analytics but actual culture to take advantage of real time streaming analytics and not just say well there goes a, we just saw a really important event just happened. Okay what do we do? We don't know the time has passed. Part of it is I'm not sure how this is all going to evolve. I think the world needs to get more, I mean there's an awful lot of us here that are drinking the Kool-Aid saying analytics is the answer. There's an awful lot of folks that don't trust analytics yet. And so part of our business strategy early on is both to use analytics that are already available that our customers are already using and David mentioned the Navy. You know we'd worked closely with the Navy for several years taking technology, both technology and algorithms that the Navy had and then adapting those into this stream highly parallel application environment that we're operating in. And so to go back to that same customer and say look we're not trying to change everything you do, what we're doing is taking what you already do and do it better. You already trust your analytics. Let's work through how we've now made those analytics better. And so I think that's a much better way to approach it rather than trying to blow into the room if you will and say analytics is the answer to somebody who didn't know they had a problem. And so I think the world's got to get more comfortable and also understand in a lot of our applications there's a trade off with false positives and all that sort of stuff. We're not making the decision, it's the customer making the decision on are we in full auto if you will. If it's a cyber attack, there isn't time to have a person in the loop so that might be full auto for example. There might be other things where what we're also working on is the visualization and the display portion of all of this because if you can't communicate it then the information if you will is kind of trapped somewhere and it's falling off the table for another reason. And so I think there's going to be an evolution and I hope our strategy of working with customers that are already having a big problem is a time critical problem and are already somewhat at least predisposed to doing analytics on that data anyway. I mean most of our customers are some of the very early, some of the first big data customers in the world. And so getting run out with them I think is a pretty good test but also as they ring it out there'll be other folks that'll kind of see it working and be a little more, you know, the classic more crossing the chasm and all that kind of stuff. So did I understand you correctly Bill? You're saying you're not spending a lot of time worrying about false positives, that's you. No, no, no, I'm saying from a technology standpoint yeah we're not worrying about that. It's really for example, we might be solving problems, we are solving problems with streams in our analytics today that literally could not be solved a year ago. And in being able to solve those it may be that we'll come up with what's they call it a figure of merit or a confidence factor to the decision maker that would say you now have a 50 or 60% confidence based on analytics that this is what's happening. What we're, in most cases or in some cases we're not going to say it's 100% because it's based on analytics. Now it's really fast and it's incredible analytics across multiple sensor streams but in a sense it's still a judgment call. And what we're trying to do is provide data and inform the decision maker and allow them to act on that however their policy describes doing it. But the capability is there to be full auto. One of the things we can do in streams and we've done and demonstrated is let's say just to make up some example you've got five different sensors available right now and there's a couple more sort of in the closet that are really expensive but you can't use those unless you have higher justification. What you can do in streaming with our analytics is you can not just do the analytics but then you can gain insight let's say in a few seconds and then based on that insight turn on those three other sensors and then bring those into the party and then raise your confidence factor based on the analytics maybe to 80% just to make up some numbers. But what we're able to do in the fabric is not just the analytics but actually take action such as turning on a video, analyzing that video then based on the video turning on some other sensor input and so you're processing and acting on all the information as it's streaming through. Can you, my last question is the data sources. Where are they coming from? How do you see them changing over time and what does that mean to you and your customers? Well the obvious thing seems to be it's all essentially some dates all going to be unstructured. I mean the numbers just are all going that way and the reason we picked InfoSphere Streams is because it was tailored, it was developed for the US government several years ago specifically to handle unstructured data and so it really doesn't run out of gas. So it could be text, could be video, could be satellite imagery, could be social media and we've demonstrated all those different kinds of applications. It could be binary up through the packets for cyber. So we're able to handle all that kind of data and so we're pretty agnostic. We're not too worried about where the world's going from a data structure standpoint because we think we've got a fabric that allows us to be able to operate on all of it. Which is a pretty incredible, yeah I mean to what you're talking about before I think IBM's done an incredible job building a great portfolio but what I keep harping on to Rob Thomas and these other guys is Streams is home built and Streams is the most incredible InfoSphere Streams has put in a plug for IBM. You know we've looked a little bit at TwitterStorm and Yahoo S4 and there's some high performance computing streaming platforms out there as well but there's still more research oriented. InfoSphere Streams is, they're announcing their 3.0 release version of it. It's prime time, it's what you'd expect if you were a business using it and what it can do is just incredible and it was developed by IBM. Well let's not forget IBM does more R&D than any company in the industry and you write on, I mean a lot of innovation actually comes out of IBM and in fact Oracle CEO Larry Ellison has made that point many times that he does worry about IBM because they do so much R&D so that was his way of backhanded slap at HP of course but it's true, IBM is renowned for its emphasis on R&D. All right Bill Hartman, Tara Eccles thanks very much for coming on theCUBE really great story, good luck with the company and the next round and good luck with your trip to Washington this week. Thanks Dan, call ahead. Keep it right here, we'll be right back live from Mandalay Bay in Las Vegas this is theCUBE, SiliconANGLE's coverage of IBM's IOD conference, keep it right there.