 And welcome back to the Think Tech Hawaii Studio. We've got an exciting episode of Security Matters for you today. Brian Kraus is with me from Dragon Fruit AI. AI in our industry is making a big impact. We're going to get into some of that technology today. We'll talk about specifically, I think, the delivery of the execution delivery that Dragon Fruit has is a little unique. Hopefully, we'll get into some of that. But first, let's meet Brian. Brian, thanks for taking the time to join me today. I think you're coming in from somewhere in the East Coast. I am. Thank you, Andrew. Great to be here. Very excited to talk with you today. I'm coming to you from the other end of the world, so to speak, up in Boston. All right on, Boston. Yeah, my buddy, Sal D'Agostino, I don't know if you run into him, crepes around the industry or not, but he's out of Boston. I've got to get up there and visit with you guys soon as soon as we can all travel again. Well, let's get into kind of your history with the industry. You know, I know we don't all give it away on social media these days, right, because you never know who's hunting you down. But, you know, as much as you'd like to share with our audience about your history and kind of how you developed in the industry and then sort of ended up with Dragon Fruit today. Sure, yeah. So my my history, I mean, I don't want to date myself too much, but I've been in the gray hair does that by itself. But I've been in the industry for, you know, in technology more broadly for about 20 years, more specifically in, you know, more software and technology geared towards the physical security world for about the last 10 or 11 years. I've always been drawn to sort of the new technologies, how to take existing investments in existing technology to the next level to do more and push the boundaries a little bit. So, you know, that being said, I've spent the last several years in the field of artificial intelligence, as it's called in the industry. Before that, I was in software defined storage for video surveillance, which was sort of cutting edge. And even before that, you know, in the in the camera manufacturer space and a mobile security startup as well. So I've had a good diverse experience in software, hardware, as well as, you know, analytics and some of the advanced tech. So it's been a really, really educational and an exciting run. Yeah, it's nice to have that sort of broad perspective and watch it over time. And then, you know, if you if you have a pension for the getting out on the edge a little bit, then, of course, AI shows up, which I think is truly finally starting to change our industry. I noticed that in your history, you had you had an MBA and you also studied to work in, I think, in finance or capital management. Having that business lens on on our industry and on the companies that you've worked with, for me, brings a different perspective. I end up talking to so many industry types that are sort of law enforcement or or a guard background or or even military background. You know, some advanced stuff comes out of the DOD, obviously, those folks. But the business folks tend to be in the in sometimes in the back room, right, the CFO types, and they'll get out. You don't meet them at the shows very often. What's your perspective on how we sort of how we sort of manage investment and money, you know, in our industry to develop things like AI and move the technologies forward that help our customers? Yeah, well, I think the most important thing and what's helped me the most and I spent the early, early days of my career in finance and investment management, I think the most important thing is to help end users and stakeholders in the physical security world understand the impact of the business that the technology has, not necessarily just how it solves for this specific use case within their environment, but how it impacts the business and how it could potentially impact the business more broadly that maybe they haven't thought about and instead of looking at security as a cost center, like traditionally, it always has helping customers to really understand how security can potentially be, you know, a revenue center or an enablement center for the business instead of just a cost center. I think that's been a big part of my success leveraging some of that background that I've had. Yeah, and I think it's not necessarily unique, but that voice doesn't get introduced often enough and early enough in sort of a solution conversation with a client. You know, we, my folks, you know, myself being military trained, you know, we're all at that layer of security approach and, you know, blah, blah, blah, blah, blah. And it looks like a cost center. It looks like an insurance policy or whatever. Maybe at that spend, spend, spend instead of that investment, how do we improve the business? And I definitely want to get in in a little bit to how I think that some of these artificial intelligence technologies are going to be able to do that and how they're doing it already for customers. And I think we've improved what I call it, not a hollow approach, but we had algorithms for years, which I'm sure you saw that work some of the time, some of the environments that you recently performed, but a lot of false positives and a lot of problems with some of that technology mostly being deployed improperly in my estimation, right, you know, but the evolution of technology where do you think we are with what you've seen? I know you came from some other AI companies prior. How do you think it's evolving? You know, we've been through the sort of machine learning piece. We've got these convolutional networks of which I have no real expertise, but it's evolving. I hear AI thrown around a lot, which to me was always sort of like machine teaching machines, and I don't know if we're really there yet, but what's your take on the technology and the buildout of it and sort of where we're at and where we're going? Yeah, it's a really good topic, you know, one that we could probably consume a whole show on. But I think generally speaking, I think we're still in the very early days. OK, good. You know, we're at that phase where there's a lot of investment coming into the space where you see a lot of people trying to develop solutions, you know, for different things. You know, we joke here at Dragon Fruit that if you go up to the roof of our CEO's house and hit a three word, you'll hit six other AI companies. So it's a very crowded space. Overall, I think that's good for the industry. I think it helps to broaden the category. It helps to broaden the awareness and get customers looking and introduced to the technology. And I think what we'll see over the next couple of years is really that consolidation start to happen. Both from a provider perspective where, you know, quality will rise to the top. You'll start to see market share taken by companies that can deliver, you know, algorithms that work in environments that already exist that don't require a lot of, you know, incremental investment to make it work. And so I think you'll also see from a technology perspective, a little bit of evolution in a sense that, OK, now we've validated the use cases in sort of the return on investment for analytics. And the next big question is where can we deliver them from? So, you know, big talk in the industry is, you know, edge core cloud and where that compute comes from. And there's a lot of work going on at the camera makers, chip makers, the cloud providers, so on and so forth. So I think we're at a very interesting place for five years from now. It's going to be a completely different industry. Analytics won't even be a topic. It'll be it'll be table stakes, so to speak, for pretty much every system out there. It'll be done differently for different different folks. But I think you'll start to see that consolidation over the next couple of years. And some of those some of those decisions being made that road map being played out. That's awesome. Yeah, we'll have that. That it's like a you won't even ask if it's there. You'll assume that these sort of features are baked in. I from my perspective, I don't recall a lot of this work or exposure. And we've had Intel show up interesting. You'll catch the hands of sort of leading that charge around the industry in the last few years. And it seems like with Intel came a lot of this work. What's your sort of take on that? I love it that they're here. I mean, we need big name like that to help our industry, you know, breach this technology gap. I just don't remember anyone else with that sort of name playing with us all these years. Yeah, yeah, yeah. Well, well, Intel is, I mean, a pioneer in the space and they've certainly helped to move things along probably more than anybody. Other names, you know, of course, like Amber Ella and NVIDIA knows folks have, of course, invested heavily and seen great returns on some of their products as well. But yeah, we need we need those folks to keep innovating, of course, because, you know, this space, you're talking about massive computational algorithms that require a lot of processing, a lot of compute and the question becomes, you know, where can we perform that compute and the smaller we can make it and the faster we can make it and the more efficient we can make it, we can put it in different places like on a camera and not have to have a server at all or in the cloud. So a lot of those things are, you know, heavy investments from the processing, the chip makers and so on and so forth. And it's really, really great for the industry. And I think, again, that that's going to lead to a lot of change over the next few years, for sure. From what you've seen on the development side, do you think I've talked to some folks about this? Do you think we'll be able to take a scenario? Let's just say we've got a camera watching a room and there's 10 people in it and we're just doing facial recognition. So the camera can handle that compute. Now, let's say the room gets 50 more people and all of a sudden that compute needs to move to a device, maybe on-prem. And now let's say there's a thousand people in the room and we've got to get that up to the cloud to process it. Is that something we're going to be able to do dynamically? Maybe in the near future, you know, where the where the app is smart up to go, ooh, I can't handle this on the camera. I've got to move where it just happens for seamlessly and does the workload. It depends on your definition of near future. I don't think it's near future. So OK, we can certainly do face detection on a camera today. There's, you know, several providers that can do that. The challenge becomes when you want to take what has been captured, right? So I've identified the face as a face. I don't know who it is yet or if it matches anybody in my database. But I've identified that there is a face here. It's gone from fifty to a thousand, as you mentioned. But now I have to actually perform that one to end comparison of what I have in my database. Do I know this person? That's the challenging piece right now that really can't be performed on the camera. And so you'll kind of, at least for that near future, again, depending on what you call near future, that will probably reside either on-prem and a server at the data center or in a cloud environment where we can perform that one to end or end to end comparison. Yeah. And it's that scaling piece I'm interested in. You know, it's you really want to be able to drop the box out there, the device, the sensor, let's call it and let it do its job, regardless of the capacity that it's faced with, right? And it's processing a thousand license plates an hour, but what if it becomes a hundred thousand license plates an hour? Like, I don't know if our traffic's that bad here, but it can be. You know, so I just don't know. I think that some of that that back end work will need to be done for. You know, all the other works getting done and it seems like pure magic, you know, to a lot of people. So the scalability of it's that next piece to be able to leverage the cloud when needed, you know, and I've been looking for somebody to go, yeah, we're doing that. But no one said that, you know, at least on camera. So we'll wait for that day. Yeah, you do with the data, right? I mean, just kind of scanning the scanning the street for license plates is is a very discreet function. Logging those license plates, comparing those license plates to databases that exist somewhere else and creating alerts and so on. That's obviously additional discreet functionality that has to live somewhere and perform somewhere. And that's where you get into the challenge of, you know, what is it you want to do with the data and how much of it can I do just out here by itself versus I need some more CPU to help it? Yeah, that's the way you try to do. You think, well, that is the investment going to be quick enough out at at the edge, you think, in these in the chipsets and the memory and the capacity that's out there? Or is it is there going to be is there a kind of a limit that's like, yeah, we're we're a ways away from, you know, that investment doesn't make enough sense yet. So let's keep it back in the box or in the cloud. Yeah, that's also a great topic. And I think that's one that the industry and customers need to be aware of. You know, I think we're at a very early stage of being able to run, you know, AI on the camera. We've got some initial chipsets that are out there from, you know, the accesses of the world, the bosses of the world's great stuff. But it's really not super advanced computing that they're doing. Right. We can do some general analytics and it's very valuable. But I think what it's creating now is sort of this excitement in the customer base that all I can just do everything on the camera. And then like we talked about here, you sort of have to have that discussion of well, yes, I can do maybe like indexing, but I can't really do anything with the data without, you know, some companion compute somewhere else. So it's great in a sense that it's, it's, you know, they're sort of pushing the envelope and creating all that awareness of what is possible. But for now, probably for the next few years, we're still going to have to bridge that gap to really provide a useful, you know, high utility, you know, product for the customer. Yeah, I love that that there's there's where it's nation. I love that we that we can talk about that being there and educate the clients in that way. We, it goes quick. So we're about halfway through. We're going to take a one minute break. We're going to pay some bills and we're going to be right back with dragon fruit and Brian Krause. Stick around. Welcome back to security matters. We're in the think tech Hawaii studio today and Brian Krause is with us remotely from Boston, Massachusetts, and we're talking dragon fruit AI. We've been past Brian's bio, so you'll have to go back and watch the recording to catch up with us there. I want to get into this technology. Dragon fruit AI. First of all, when did you guys launch? How long have you been around? Yeah, the company's been around for about a couple of years now. You know, like, like many companies, we spent the first year or so building, you know, version one of the product and now we're sort of in that commercial stage where we started launching commercially last year. And now we're in sort of iteration phase refining and adding to the product based on customer feedback is very exciting. That's awesome. Yeah, I went to them, the use cases, like a module graph, which was really important to me. I was like, wow, here's, here's a great way to look at a vertical market and say, here's the things that are work. And I'm presuming those sort of came from customer requests or, or something like that. And, you know, most people develop where the need is. Um, and I thought, wow, now this is awesome. But I saw that some of the deliverables are even by the hour. So I want to, I want to get to the deliverable, but I'd like to talk a little bit about the execution piece when you, when you're building this. So you have to either train or teach an algorithm to learn or, or AI to learn something about an environment. And my understanding is that's done sort of en masse. But then when you get down to the customer site, you have to to refine that a little bit. Is that still the approach or is that, is that different from the approach dragon for each taking? Uh, it's a great question. So just to kind of clarify, um, so there's, there's machine learning and then there's deep learning. Um, deep learning obviously is a subset of machine learning. Um, and, and what, what we and what most analytic providers do is we're, we're not using the customer's data to train our, our network. Okay. Um, so when we deploy, you get the latest, greatest version of our algorithms and obviously being cod based, we're pushing those down, uh, you know, into a production environment, you know, to push up a button. Uh, which is fantastic. Um, but we're not, we're not refining our models based on your employees behaviors or anything like that. Got you. Elsewhere. Um, so there's, there's no deep learning going on from the customer, customer video, uh, but they are getting the latest and greatest algorithms every time they log in. This is one of the benefits of, you know, being a SaaS provider. And, and that algorithm is continually improved from the learning it's doing in real world environments, I'm guessing. So it's constantly improving. Yeah. The algorithms are always tested and improved on, you know, based on data sets, right? So we develop an algorithm and then just continue to, uh, feed it more for lack of a better term, um, just refine those software algorithms to become more crisp, um, and being able to make decisions based on the type of data that we see from physical security networks so that they can make the decisions that they need to make. Um, it's not always perfect, as you know, in a security environment, the, the images that we get and some of the data we get. So you really have to have sharp models that are constantly trained up. Sure. That's awesome. Um, I was looking at just the, the use case module there. And so, and there's, I see a lot in retail and then obviously law enforcement work, venue work, um, smart and state cities, which I think has got a lot of use case stuff. All the folks in New Orleans are, are using this stuff, taking cities, surveillance cameras that are, so we're sort of placed for security, but able to alert on flooding. For example, uh, my teachers look for flooding in certain areas of the city, things like that, and alert people to go, I guess, turn on the pumps or whatever they need. They got water problems there. Um, what, um, what do you think the specific sort of, um, um, use cases that have jumped out at you folks that had like your early success on or where you, you know, you've gone in? I know retail has been big for everybody. Is that sort of where you got started or was it something else? Um, it is one of our early verticals. I think, you know, with our product, we really, um, we really focus on, uh, two key areas. One is what we call occupancy management, um, which is, which is all the analytics that relate to the operations of physical space. So, you know, I want to have a better understanding. I want to drive intelligence out of what's happening in my store, in my casino, in my business, on my manufacturing floor, et cetera. Um, so we, we've got a whole suite of analytics that are designed just for that. Um, and then the other half of our business really is, um, in the post video review, which is huge towards law enforcement, um, security heavy users, which is I've got an incident, um, I've got tons of video that I have to pour through and I need a faster, more efficient way to do that because particularly in the last 12 or 15 months with COVID, um, we are now in do more with less on steroids. Um, and so everybody is looking for tools of efficiency. So early on, those are the real two key areas. Of, you know, is my space being utilized? Do I really need this space? How do I optimize this investment in physical space? And then how do I automate processes of speeding things up when something happened? Um, those are the generally the two broad areas that we've, we've got out early on. And are the consumers, so when that, if you cross those, obviously the investigation guys are typically probably operators, law enforcement operators, or, um, I can understand the investigative types, you know, from an inside of a corporate, you know, Fortune 500 firm, whatever it may be, um, what about the occupancy types? Does that, because to me, that, that's, you know, kind of gets into manufacturing floor or distribution center, um, you know, uh, even, you know, occupancy itself, I remember being at, uh, oh, I don't know, Cisco back maybe 15 years ago, and they were looking at that for to shrink in their campus size, you know, where they had a building that was, you know, 20% utilized and another one that was 60%, they said, Hey, let's move all those people over and shut that one building down, you know, on their campus. Um, is that kind of stuff happening like in the real estate industry, or, or what, what do you kind of see in that occupancy peak? Absolutely. Absolutely. Yeah. You know, it's, we definitely learned a lot in that space when it comes to corporate real estate. And, and one of the most interesting things is that some of the states actually tax these corporations based on how much their space is actually used. Wow. Okay. So, you know, I've worked with some companies that have offices in California, for example, and they've gone through, you know, big audits or I don't know, challenges, if you want to call it with the IRS to say, you know, look at we've, we've actually put these systems in place to prove to you that our tax bill for, you know, the usage of this property is lower than what you're charging us. And we've saved, you know, two, three, four million dollars in single years. Wow. So it's, it's very powerful and, and, you know, obviously with, with COVID and everything else going on in the world, obviously everybody is looking at occupancy. You know, some folks are still doing it. I mean, in retail, you know, I was at, I was at a store this past weekend and I'm still seeing people with an iPad at the front door doing one in one out, like, like at the bar, you know, when you're in college. And to me, it's, it's, it's a little bit crazy that we're still doing that manually when, when I look up and I see a camera right above the person. And I said, you know, I say, you know, that person can go to work and do their job and we could automate that entire process tomorrow. Yeah. And, and so we're seeing a lot of focus on, on that. It really speaks to, again, sort of more broadly, like to do more with last, but we've got a lot of groups within companies, like take retail, for example, where lost prevention, where they never had to worry about any of this stuff. Um, you know, they were focused on shrinkage and the tools and the processes to stop shrinkage. Now they have to worry about social distancing, mass compliance, occupancy of the stores, et cetera. And so at the same time, the team is the half the size it used to be, right? Because top line revenues. So you really are put in a corner to find solutions that can automate and offload a lot of the work that used to be manual, some of which is new. And that's exactly where we live, which is, you know, if you're faced with these types of challenges, let's take a look and see if we can, we can automate a lot of that stuff. So you don't have to, you don't have to put a drag on the team and, um, and not be able to accomplish it. Yeah. Like the one in one out, I think we solved that for parking a lot of years ago, right? You show up and there's a sign. There's 300 empty stalls. You're good to go. I think they have that in my hospital now. So like to not solve it for people, especially now. Well, I mean, we've solved it if people haven't deployed it, but let's talk about the deployment a little bit. So folks engage, I saw that you had, you had, um, you actually offer certain, um, uh, algorithms or analytics for like by the hour and things like that. I thought it was amazing. So, um, talk, talk about the delivery model and how people can engage, um, you know, with dragon fruit and take advantage of some of the technology. Yeah. Our delivery model is one of our biggest differentiators. So, you know, I, I, we came here and started this company to kind of democratize AI, if you will. And when you look at the, the landscape of traditional physical security analytics, it's largely software license base, license, the camera, um, and for customers that have 5,000 cameras, it's, it's a non-starter, um, because I, you know, we just, we can't afford to put a software license on every camera and then put a huge compute environment in place. So what we do is, you know, we have a very flexible modular driven platform that says, if you just want to do people counting at the front door of your store, um, on that one camera in your 50 locations, then that's all you need to purchase and that's all you need to consume. And if, you know, you want to do some more things later on, you just turn them on. And if COVID goes away and you no longer have to worry about occupancy, you can sunset that functionality and use some other capabilities in the platform. And it's, you know, very tight to your return on investment model. You know, you don't have to over subscribe or over invest into a big system and use 10% of it, right? It's similar to like what's going on the consumer side where, you know, you have the cable bundle versus the streaming choice. You know, I want to get rid of my cable bundle because I don't need to pay all that because I don't use it. I don't watch it. And I can get by with Netflix and Hulu and maybe, you know, one or two other things and save myself 50% per month. That's the model can be applied to enterprise when it comes to analytics, which is we can right size exactly what you're trying to do with the software. And we can, and it can move over time if you have to. And that's very exciting to be able to bring to the market. It's a big differentiator. It gets us a lot of conversation. Yeah, I think it's super important. It's kind of like, you know, the way Microsoft bundles 0365, right? You can turn it on. You can turn it off. Those may be monthly, you know, for a company, but like in the government space, that's not, you got to buy about a year, which is still like a real problem for my government network, in my opinion. So I love the right sizing. I love that, that delivery, like that kind of use it as you go, consume it as you need. You could spin it up for an event in a certain area for, you know, as long as the cameras are there, right? Which is super flexible. What's the sort of, for the customers, we only got a couple of minute stuff, for your, for your customers, how's that engaging? Is it remote for them? They got to pick up a phone. Do they turn this on themselves and then turn it back off themselves? How's that work? Yeah, so we do have a free trial on our site. So if you go to Dragonfly, you can, there's a little magenta button in the top right, and we'll be doing our site. So if you're watching this after the end of April, it might be different location on the site, but it'll stay try for free. And you can sign up and we put some demo content in there, and you can, you can add some videos from your own environment to see how the AI works. But if you're, if you're looking to do something a little more complex than that, I guess, you know, in an enterprise space, you want to, you want to work with us and we have engineers that can support you in through testing and evaluation. You know, you can reach out to us at any time. I'm at Brian at dragonfruit.ai. And, you know, we have info at dragonfruit.ai. We've got the Black Channel in support on the pre-trial. So if you need to get all of us, it shouldn't be too difficult. But we've got a number of ways to engage with us and we're happy to work with, you know, any customer on what they're trying to solve and try to help them look at our technology to do it. I love this, Brian. Thank you so much. Hey, out there in the world, if you want to get engaged with some AI, it doesn't get a whole lot simpler than what dragonfruit is bringing to you. Take it for a test drive, dragonfruit.ai, or get Brian a call, get engaged because this is the world as we know it sooner than we know it, I think. Brian, thanks again so much for joining me today. Have a great rest of your week. Aloha, everybody. Take care out there. Thanks, Andrew.