Ubiety | IPVM profile & technical discussion

6:29PM Feb 18, 2022

Speakers:

Keywords:

sensor

cellular

home

people

devices

data

detection

house

network

wi fi network

nikita

company

wi fi

mobile app

hardware

detect

insights

carrier

phone

deploying

Good morning. Good morning. It's afternoon for us, I guess. Well, yeah. Afternoon for me too. But just y'all are east coast, I guess. Yeah, yeah. Yeah.

Yeah. He's from New York.

Nice. Nice. How's the weather over there? It's chilly where we are. Oh, yeah, it

started out as like 62 degrees in the middle of the night now. It's

like 23 big shifts.

By the way, can you hear me?

Nikita you're muted?

Is the obligatory mic MIC CHECK part of the meeting that happened

to me Yeah. When you spend like two minutes can you hear me now? No, no. We can't hear You. Trying to learn more every day. So that's why we were doing calls like that and talk to companies like yours.

Okay, let's try this one. Can you Mmm, cool. Thank you. Sorry for the for the confusion. Sorry about that. So Mike, nice to meet you on Nikita. I'll be joining. I'm joined by Ilya today, who is also Research Engineer at our company. So I guess we should start with the introduction of who IBM is and who we are. So IBM is a independent testing and research company that was initially centered around video surveillance. But now we like we are broadening our horizons and we going to access controls systems, robotics, weapons detections and intrusion detection and obviously, AI and stuff will wait here and other companies so we actually researching them and doing reports and test them if we can. So yeah, my name is Nikita. I'm from Moscow, Russia. I got my undergrad degree there and most prestigious university in physics. And last year, I moved here to pursue my master's degree, which I successfully got from El and now it's been almost four months, and I've been working the DACA game here. So Ilya also has a physics background from Penn State, if I'm correct,

yep. Yes. My educational background is not as impressive as the Ketos was, yeah, I graduated from Penn State last May with a degree in electromechanical engineer. Nice.

Yes. So um, by the way, if are you okay with me recording the audio of this conversation? So it just will be easier for us to cite your Okay, cool. Thank you. Yeah, that's that. So can you give us a like a broad description of what your BI t is and what do you and your company and like, the perspectives?

Sure. You all were kind enough to give me quick backgrounds about yourselves. I'll do the same. My name is Mick Cox. I'm the Chief Technology Officer here at bidi party of it. I was only the NASA Jet Propulsion Lab down in Pasadena, where I went there Internet of Things team. So we were in a group that was directly under their Chief Technology and Innovation Officer and the charter was to find new technologies that are going to be impactful over the next decade or so, and actually prototype them, bring them in into the lab and kind of work as internal consultants to sell them to all the really smart engineers and scientists who are supposed to put rovers on Mars, right? And once you learn how to land a rover on Mars, you don't really change it a lot of times and so our charter was to bring in new technologies, show the benefits of them and kind of help help sell through prototyping. So I got to play with a lot of new technologies, cutting edge stuff. JPL sits on the customer advisory board for Amazon Web Services. And they have for a couple of decades now. And as a byproduct of that. Whenever AWS was coming out with new services, and a lot of times, especially in the IoT space, it would actually bring those to me and my team first to kind of cut our teeth on and play with and tell them what we liked about them and what we did. And so as a byproduct of that I got to play with a lot of the really early alphas for some of the Amazon web services that are now kind of public service offerings and spent some time on stage with their VPS announcing some of those new services that were coming out and you know, including in natural language processing was somebody Alexa and Lex stuff, robotics with some of the robot maker work that they were doing IoT with Greengrass and some of that work and just just an amazing an amazing time. I really had my dream big over at JPL. Like I got to work on whatever technology I wanted. It was all new stuff. Really, really cutting edge stuff. Some of it awesome, some of it not great. It was high level of autonomy. The NASA name is obviously really cool. But when when our co founders Keith bucket and rich SB approached me about what you buy it he was working on. I kind of went, Oh, wait a minute. This could be really important and they describe to me the problem. And I started doing kind of five hours a week on the side turned into 10 hours nights and weekends turned into 20 hours turned into 40 hours alongside the JPL stuff. And then eventually finally made sense just to make the full time switch. So I was brought on in in a technical role to build the IoT platform that powers our sensor fleet, and all of the sensor detection code that runs on our fleet of sensors. So that was the original author of some of that stuff. My role has kind of changed as the company has grown and morphed into what it is today. So trying to keep the technology pointed in an amazing direction and let the people that are smarter than me the engineers to do incredible work. So that's kind of where we are. You buy it as a really high level. We are really trying to create what we call presence fidelity, using RF devices as a proxy. So that basically can tell us what devices are in an area of interest that emit RF, whether that's cell phones or headphones or laptops or watches smartwatches keyboards, whatever. And there's a high degree of, of, of correlation between the devices that are in a place and the people that are in the place as well. I can't remember the last time I left home without my smartphone. And so if somebody could detect that my smartphone was in an area, that's a pretty good indication that I'm in the area as well. So we started with with that kind of idea in mind and we were primarily focused at first we're out of Chicago. Chicago has a pretty serious gun violence problem. It's getting better arguably, but it's still something that we have some room to improve on. And so Chicago and other cities have employed sort of IoT detection sensor fleets, in some different regards for different cities, mostly around video data or audio data. So for example, if you if you're familiar with like a ShotSpotter kind of a system. This is an array of microphones that does gunshot detection and then can train cameras to go look, you know, in an area and while the idea behind ShotSpotter is fantastic. I love what what ShotSpotter has done. We found from law enforcement having done some ride alongs with Chicago PD, Cincinnati, PD, LAPD, the data that systems like ShotSpotter produce tends to be sort of repetitive, you kind of get the same types of videos over and over and over. And a lot of times that doesn't result in as much actionable intelligence as you would want it to. You see kind of a fuzzy figure at night. There's not really that much you can prosecute based on just kind of having seen the video of somebody wearing a hood. And so we realized if there was a way for us on the cellular side, to detect the subscriber ID or a phone that was in an area, if a violent crime were to occur, we could take the data of the phones that were nearby through a subpoena process and actually figure out the personal identity of the folks that were around and the phones that were in that area. So that was the first the first dream that we had was to augment something like ShotSpotter to say okay, you know, you're great at detecting gunshots, you're great at video recording. We want to be an Augment to kind of give you some of that that RF detection as well now to tell you what devices were in the area at that time, to hopefully give you a little bit more prosecutable and again, objective and unbiased information to be able to actually go accurately prosecute those crimes. So we started kind of thinking about that from a city safety lens. We've sort of grown as our as our company has grown in not just looking at it from a city safety, but also thinking about commercial security applications. If you think about, you know, businesses or retail, there have been a lot of smashing grabs in Chicago and other you know, other places of course, across the country as well. Many of those have struggled to do things like prosecute the crimes and even simple things like count the number of people that that broke into your your, you know, into your facility. And so what we're hoping is that on the commercial side, our technology could also help detect how many devices were in the area to get people more useful, actionable, information when you're filling out an insurance report and that kind of thing, as well as actually prosecute the crime. So that's on the commercial side. And then on the home side, we realized if you can detect the devices that are coming and going from an area of interest, that could be very, very powerful in a residential scenario. where not only security is interesting, but just the awareness side of that is interesting, too, right? If I can answer questions like hey, did my kids get home safely at the end of the day after school? Did the housekeeper stay for the full two hours that they were supposed to be there? Hey, I'm away from my home. For The Weekend is my teenager throwing a party with 55 kids in the house? Like all of those kinds of questions, we realized this is something that we could answer with our technology as well. And so we spent some time on the residential side as well. And that's kind of the you'll you'll hear you buy it which is the name of our company. You'll also hear homeware, which is the first product that we're looking to push out to the market. homeware.com is a website that has all the product information and then you find e.io is the website for for kind of the company as a whole. So that's a lot of ways a lot of words to say we can detect devices that are in the area. Based on Wi Fi, Bluetooth and cellular and the information that they they emit. Our sensor is completely passive. So we don't, there are a lot of sort of, you know, three letter agencies that can black hole your device and pretend to be a cell tower and kind of read the messages and all that kind of stuff. We don't do any of that. We only look at the signaling that your device is emitting just to get either to the Wi Fi router or to the cell tower. And we can detect uniqueness based on that signaling. And that's all we're trying to do. So So some things of interest that we get.

There's sort of four main datasets that we're really, really interested in that we're focused on today. The first one is on Wi Fi network detection. So our we have a physical, I can actually show it and we have a little physical sensor box. We've got all of the all of these things. They're renderings and all that kind of stuff on the website. Do but we have a physical sensor. There's a couple of different models of it. And then we have a mobile app today. And what we do with that is the homeowner installs the sensor, they plug it into power, and then you just go on to the mobile app and get it connected to your Wi Fi network. And that's all you need to do. But once it's on the Wi Fi network, that first data set that we're interested in, we can actually look around on the Wi Fi network and get a lot of information about the devices that are connected to the Wi Fi network. So we can tell things like hey, it looks like your Xbox is online. It looks like you got six different Apple TVs that are plugged in. You've got five different you know, Apple iPhones are on here. And then here's some information about those iPhones. Through the mobile app. People were able to kind of go through and say yep, that's my iPhone. That's that's, you know, Nikitas iPhone, this is Ilias iPhone. And then as soon as those iPhones come and go on the network, you can actually see that in your mobile application so you can see when people are coming and going from your house so that works really well for the residential, and for the non residents, right. The people that live in the house, they're almost always connected to the Wi Fi network. And so it just is a really, really good proxy for that. So that's the first data set we're interested in. The second data set is a little bit different than that. It's what we call sort of passive Wi Fi. So your device, even if it's not connected to a Wi Fi network is still emitting packets via Wi Fi, usually to try and figure out what Wi Fi networks are around it that it can connect to. So for example, you know, MCs, apartment Wi Fi my phone needs to know whether that Wi Fi network is there so that it can try and automatically connect to it. And so periodically, what phones will admit is hey, is MCs apartment Wi Fi there and our sensor can actually listen for those and detect it. And so we can see a list of some of the Wi Fi networks that devices have connected to in the past.

So it's less on the past day alive. So it's like stay life identification. Are you there? You're still there, right?

It's a it's more discovery than the stay alive side. So if I was you know, I'm walking around the city and I come back home for my phone to figure out oh look makes apartment Wi Fi is here. I need to connect to that right. It's kind of more for that automatic joining and discovery scenario. But one of the interesting things about that is that like for example at our we have we lease a little demo house, out in the in the suburbs, the single family home where we install all this equipment and do kind of research and testing. In demo house, our sensor picked up on here, the JPL IoT network, which is actually a hidden Wi Fi network that I haven't connected to in probably three or four years. But it's still something that my phone occasionally emits out asking whether or not it's there. And so if I were to jump into a into a house and steal something, and my phone in, you know, a homework sensor were to see JPL IoT network, like that's a that's a pretty good clue. Right? That's a pretty good clue. So that's that's the kind of data that we're interested in on the passive Wi Fi side. The third dataset is passive Bluetooth. You can imagine, you know, for phones to connect to air pods and all the other things, you have to advertise your presence via Bluetooth. And a lot of times devices actually divulge a lot of information about what they are and who owns them because typically people name your iPhone makitas iPhone or whatever it is. And the phones will often advertise that or sensor just as paying attention to those advertising packets that the the Bluetooth is putting off to kind of augment all the other data that we're collecting on the other side. So those are the three and then the last one that's really important to us is the cellular side. And on cellular again all passive we have a software defined radio based approach where in LTE, especially we can listen to the heartbeats and the handshake messages when your phone is trying to initiate a connection with a cell tower. So in order to do that, the phone has to say, Hey, this is a version of my subscriber ID you should let me connect to your network and then there's a negotiation that happens back and forth between the tower and the phone. We can actually sit there and listen to that negotiation and say, oh look, subscriber ID number 1234567 Looks like it's in this area at this time. From that we also get information about how many phones are there at a high level proxy. So thing and and the carriers that they're on and the frequencies that they're communicating on. So as an example, if Nikita your verizon family, let's say you've got you know, five people at home with a Verizon cell phone, if suddenly we see a TN T activity in your house at two in the morning. That's an alarm, right? That's an alert. This is something new that is different. And so that's that's the kind of information that we get on the cellular side. The last piece that's really interesting on the cellular side to us, is that subscriber ID that I'm kind of talking about, like when I go to the Verizon or the at&t store and link up an account or start an account, they handed me a Subscriber Identity 1234 My phone is aware of that Subscriber Identity. It's called sort of your MZ IMSI. That's your International Mobile Subscriber Identity. And as in certain scenarios, your MZ is actually what is being divulged and what our sensor can detect. Obviously, there are privacy implications around always transmitting your MZ because theoretically, people could just track you with that and that was an issue with 2g and 3g and 4g, they implemented something called the timsy, which is a temporary mobile subscriber ID which is a rotating version of that. And even though it rotates, and those are the things that we're picking up, we've done a lot of studies on how long tins these last four different network carrier pairs and in different geographies, and you can take that timsy in a specific place in a specific time. It's still mapped by the carrier one to one to that Subscriber Identity. And so if we have a list of the Tim Z's that were in a house during the scene of a crime, we can still you know, through a subpoena process get that information directly back and get that personally identifiable information. So on cellular that's kind of the like really meaty, forensically valuable data set for us. And that's why we've spent a number of years developing this sensor that I'm holding, which is that software defined radio based approach and we've got, in addition to just the amount of time and engineering resources and ingenuity that went into this, we also do have now four patents granted against that cellular detection capability as well. So we think we're really well positioned on that side.

Yeah, I actually reviewed one of your buttons and yeah, there was a very big discussion of Team Z and Z MCS and other like identity identifications. So just to reiterate what you said on the on the seller side, because this was the main like, the baton went on and on about the salary deduction. So basically, once the our salary wise our cell phones from time to time send stay alive signals to the nearest cell towers. So they send attach request messages and then receive the identity request and attention response and out and you just and your senses. Your senses just can listen to these like messages and to and decode them in order to get the important forensic details about do is is that accurate?

Yeah, I love well that's exactly what we're doing. We also in addition to what you just said, you can think about the the process of initiating a connection. So if my phone is sitting idle in my pocket for 30 seconds or a minute, and then I receive an email or whatever it is, my phone will actually reach back out and try and re initiate a connection if it had gone idle in the tower has a way to kind of initiate that connection to. And so we're actually listening in addition to what you mentioned, to what's called the Ratch sequence, random access control header and that handshake that goes back and forth. And we decode that entire stack all the way up to sort of where it was encrypted. And there's a lot of interesting sort of metadata in that in that first handshake.

And also, the difficult thing about the discovery on cellular networks is you know, different cell different carriers, different cell phones, different towers use different frequencies. Time offsets and other like identifications and distinguishing capabilities. Can you clarify how sensors can actually understand what frequency you should listen to? And what time of said should I apply to in order to to decode the information that is hidden within the messages and hatchings?

Sure, yeah, very, very good question. So the the reason that we went with a software defined radio based approach was that we knew we had to build something modular, that we would be able to tweak and change as the spec evolved, right. In addition, to that, the hardware that we have in here mimics basically eight software defined radios simultaneously. So you can imagine like that has a lot of very interesting action capability. In addition to the eight software defined radios we also have for Wi Fi radios and for Bluetooth radio. So that's a really, really capable detection device that we've come up with. But more interesting than that is the way that this sensor has been architected to have those multiple software defined radios at the same time, or operational at the same time. Is we can actually decode on multiple carriers. When I say carrier, at&t, Verizon, T Mobile sprint, we're going to decode on multiple carriers simultaneously, because we have multiple radio resources that are at our disposal. And so even though each of those individual carriers tends to use different configurations, and then in addition to that, not only do they change per carrier, but they actually changed per cell tower in some cases to part of the Ratch sequence that is happening or even slightly before the rash sequence is something called a sim and amid so master information block in a system information block, it's part of the LTE spec. Towers are emitting those, and we can sit back and decode those and listen to see how the towers are configured to communicate. That gives us sort of the realm of the possible for where a cell phone could be. And that helps us kind of narrow down where we actually need to look. The second step to that is, we look sort of for what I would call uplink activity. So now that we hear from the towers, how they're configured to receive activity, we narrow down our search and now we look around us. Are there any phones that look or is there any, you know, blips in activity around frequencies? We laser focus in on that even further, and now we can say, oh, yeah, this looked exactly like an LTE preamble or whatever it is. And so we're doing kind of that multi step process by listening to the downlink. By listening to if there's activity on the uplink. And then we know exactly where it will be because of the configuration that we heard from the tower, and we can park radio resources to listen to the uplink frequency and the downlink frequency

Cool. Cool. So within this pattern that I've read, there is different examples of app interfaces. And for example, there was a screenshot of the app when when it says like someone has been detected and number of details was written into the like this small window or whatever. So it had the name of the device DMC carrier and last seen time. So based on the I understand that pay the patent is a is not very rigorous thing in terms of like, you guys evolve your product and your technology. evolves and patent may go sleep and based on that I saw the marketing video on the OBYT website and it has much more sophisticated details about the device including like name network hostname manufacturers sell your cell phone number, which is also interesting and previous Wi Fi as you mentioned, the JBL IoT network, and also MAC address. Can you clarify what at what stage does the app like at what stage is the app works? What What details should I expect in order to

Oh, sure. So the the fields that are on the marketing video, are a combination of all of the data sets that we've talked about, which are on network passive, Wi Fi, passive Bluetooth and cellular. So presumably in the real world, you'll probably get some subset of those from an actual device. But if you were to receive all of them, you could you know, we're folding all the data together into that view that you would you would kind of see on the on that marketing screenshot. I could show you it will take too long maybe. But what we're what we're doing right now, we're in a closed beta with our mobile app, because we've actually sort of scoped that down to we want to be really, really accurate about focusing on just the on network piece at first, because that gives you the home awareness information. And so what the app does today, like the marketing piece, you know, that's where we're going. What it does today is for anybody who comes and goes from your house, that is sort of a known resident that would be connected to your Wi Fi. We're showing when those people are coming and going and that's what we do. today. We're also in addition to that, with the sensor hardware that we have, we're collecting a lot of data on the back end on the other three datasets that we're still kind of sorting through and making sense of right. And so on the passive Wi Fi side and the passive Bluetooth side. We could do things like you know, it looks like there are 35 different devices in here that aren't connected to your Wi Fi network, right, that's probably an alert. And so we're still baking out some of those kind of I would call them like pieces of insight to roll them together. We've done now three different versions of our mobile app. The first one was kind of just very prototype to show everybody what it could do. The second one was I'm going to show you all of the raw data that I'm seeing from your sensor and we rolled that out in a closed beta. And overwhelmingly, people said this is too much data. I don't know what to do with any of this. And all these pings are like this. This is not useful to me. And so we took a lot of the feedback that they were giving. And what it came down to was I need this to be simplified down to the things that I care about, which are basically the people and usually the mobile phones, right. So so the third version of the app today has a laser focus on mobile phones and on things that are attached to your Wi Fi network. And so that's what we're doing today in the mobile app is once you get your mobile phone, you know attached to you usually we see the host name and it would say something like Nikitas iPhone, so that's pretty easy to link up. Once we get that attached, you can do all of the known resident coming and going kind of flow without even touching into the passive Wi Fi the passive Bluetooth or the cellular data. What we're super interested in though, well, so I'll take this will be a two phase answer. The first phase is on the on network data alone. We have seen some pretty immense and surprising value not just to home security players, but also to for example, monitoring centers. So if you think about a monitoring center that is getting constant calls from ADT, or SimpliSafe, or whoever it is, you know, as an example back in December, we signed our first commercial commercial partnership with.

Were telling us was Hey 97% of the time we get an alert from from my house, it's a false alarm. We don't actually need to do anything or there's not actually anything going on 97% of the time. And every single time one of those alerts comes in, somebody has to pick up the phone and call somebody and verify and do do this whole kind of, you know, this whole flow. And so what they said to us was this passive Wi Fi passive Bluetooth cellular stuff. That's awesome whenever you guys get there, but in the meantime this on network stuff could change the game for us. Because if you can tell me when an alert comes in which known residents are in the house and which ones aren't. I can change my whole flow and how I respond to that. So as an example, like there's motion in the driver, the back doors open. That doesn't tell me much context about what's actually going on. But if I get an alert that says the back door is open, plus the homeware data that says there's no known resident in the house, now we score that alarm much higher. And I can actually deal with that quickly. Versus if there's maybe you know, whoever set up the account is the account owner. Number two on the call list is the spouse number three on the call list is a grandparent number four is an uncle like then a kid then the sister whatever. You've got this call list of people you know six people long if you get an alarm, and the top five of those aren't there. You shouldn't waste your time calling those people because they don't know anything about what's actually going on in the house. And so what rapid is talking to us about is saying, Hey, we've got these call lists. You tell us which of those people are actually in the house or not. And we can you know, short circuit, this whole thing, it's gonna save us a ton of time and money as a monitoring center, because I can directly call the right person. And to them obviously, that's super important because that's directly their bottom line. They're, you know, they have to pay people to sit and answer the phones and call phones and all that kind of stuff. So so it's been really interesting just on the on network side. The second piece of that that we were surprised by was sort of on this, what we kind of call the data exhaust of what we're creating, which is in order to figure out when things are on the network and when they're not. Like I mentioned earlier, we can tell how many Apple TVs are in a house when the Xbox is on when it's not, you know, how many laptops are there all of that kind of thing. We're actually having some conversations with home insurance providers, who said to us, that data set is immensely valuable, because if if home insurance providers were to know how many Apple TVs are in a house, that changes the risk profile of that home, and you can price a policy more accurately based on that subset of data. So understanding when devices are there and when they're not really, really important. The second thing that they talked about that we hadn't even considered was many, you know, pricing the policy is important, but making sure that we don't have a payout is most important, right? And most of the payouts don't actually come from security issues that come from fire smoke or water and water being the primary of that. And so what they said to us was many companies you know, themselves included, have servers that they put throughout homes that are connected to the Wi Fi network. These things are battery powered, and a lot of times the batteries go out and they die. People don't know when they're water sensors are out of battery A lot of times and they said to us if you can watch on the network for when that water sensor is there and when it's not. You could alert the homeowner and say hey, it's time to change your battery on this, you know, on this water detector. They were also interested in this from saying if if that water sensor is connected, and if it goes off like oh no, there's a water leak in the house. If I look at your data set and say which of the known residents are home, I can call number six on the list because they're actually there I can say hey, there's a water leak downstairs. You need to take a look at this immediately and close that loop much faster than trying to call the account owner who's on business travel and the spouse who's with them and all this other stuff. So we've been we've been surprised by the level of interest from just the on network information. And again, you can get that from our sensor without needing to install new router equipment or Wi Fi equipment which is good. And then the second prong is on the Wi Fi the Bluetooth and the cellular side the passive.

There's there's immense interest on on that side alone, especially because of the forensics value that it can provide which is kind of what we talked about earlier. If you can get something from that data obviously you know not everybody who robbed your home is going to be connected to your Wi Fi network. And so if you can get that cellular information that's actually the loopback directly one to one to a person to do that to be able to actually go prosecute that crime. In addition to that, there's sort of the the awareness value that that can give as well. Think of a scenario where you're at a town or your kids throwing a party, if we can tell that there 65 iPhones in your house, even if they're not connected to the Wi Fi. That's an alert situation, right? And so that's what that passive Wi Fi data set kind of kind of enables. On those three, especially you know, the on network one a little bit but then the three, especially much of the IP that we have generated over the last couple of years and most of what we have been heads down building is the artificial intelligence engine that we need on the back end. In order to make sense of all of the data that comes off of the sensors. You can imagine, like you, we have so many different devices and there's only going to be more and more of them and they emit so much information and there's only ever going to be more of that information over time. And so you're getting this sort of wall of noise of RF data on Wi Fi network passive Wi Fi passive Bluetooth cellular you're getting all this information at a really high velocity. It's really difficult to make sense of that information. And so we've actually developed an internal artificial intelligence engine we've codenamed it Eckleburg I don't know if you're a Great Gatsby fan. But the Eckleburg references is a good one. There. So we've codenamed our AI engine Eckleburg. And it basically what its job is ingest that high velocity information on cellular Wi Fi Bluetooth and on network and give us high level insights about how many devices are actually in the house. And this was not an easy thing to build. As an example, one of the things we had to do to get to a place where this was functional was we've actually purchased a fleet of Faraday cages which are these big metal boxes as a physicist you know this I got a physics background. Big metal boxes, no signal goes in or out right and so we have a fleet of our sensors inside Faraday cages. And we actually rotate out phones, mobile phones into each of those Faraday cages and we do isolation runs on each of those mobile phones. And we do that during different types of activity on the phone. So if the phone is idle if the phones in low power mode, different operating system versions, so we're actually collecting really clean datasets for what it looks like, for example, for an iPhone eight running iOS 12 in low battery mode, like we have that snapshot and we know what that fingerprint looks like. So we've actually built a catalogue of these device signaling fingerprints that we've been able to use internally to train these machine learning models so that when we go out to the general world, and we see this wall of noise, we can say basically, and this is what Everberg is helping us do. Hey it looks a lot like there's you know probably three different iPhones here and iPhone eight and iPhone nine and my iPhone 12 Because the the patterns that I'm seeing look like a combination of those individual fingerprints that I've seen in the lab before. And so much of the IP that we're building and the the brains of the team has been around creating this high velocity artificial intelligence engine that elevates those high level insights based on the wall of noise that is coming in. And kind of where we are in our journey on that one you asked specifically about the mobile app. We have we kind of have this it's internal. There's this like pyramid of kind of insights that we're building towards Eckleburg. And on that, that bottom and second layers are things like, you know how many Apple devices are there here? Right? How many how many Samsung devices are here. How many mobile phones are here, right and that's kind of the basic level but as soon as you take that up a notch can be okay. Does this iPhone show up at a similar time as this apple watch all the time? And now you can start to draw kind of correlations between those two devices that usually come and go at the same time. Right and so you kind of build on these layers. And then the top layer of the pyramid is kind of the ultimate goal, which is okay, exactly who's in the house now based on all of the combination of the devices that I saw coming and going, and does that does that correlate with you know, actual the actual real life situation that is occurring in the house? And so those are kind of the pyramids that we're building? I would say we're, we've built there's a lot of engineering work that needed to happen to build sort of the base layer. So we've, we've gotten through most of that, and we're really starting to build out kind of that first and second layer of insights. Now to elevate things into the mobile app. So as an example, this dad has a lot of words, but I'll try and give you the soundbite example here. As an example of one of those insights in our mobile app today. If there is a type of phone that we have not ever seen before, and if there is no known residence in your house, and we see a new type of mobile phone go in, you get a push notification. And that's basically saying like, okay, you know, Nikita, you've you've got iPhone eight, nine and 10 in your house all the time. If you're gone and I see a Samsung S 20 This is probably gonna work for you. And so that's kind of the first level of learning that we've built into the app. We're constantly working to build more and more of those things in. And as the data team continues to get better insights onto, you know, into the actual raw data, we're able to build better insights over time from that engine does that make sense? Yeah, words, but

that's definitely a lot of words. But thank you very much. It was yeah, it was very interesting. And especially the training your AI engine is is really is really cool. I think. So yeah. You've mentioned that you you, you showed us the one of your products, which is V three F if I'm correct. So you have two products V three F and V V three M, which the main distinction is that we three F can do seller deduction, and V three. And so, you mentioned that they both in an event in the betta testing procedures. Can you elaborate on what's the what's this process? How many sites do does, how do you tested what the success rate when we are? When should we expect like the commercial release and that kind of stuff. So better released better testing really?

Yeah. So there's, there's you mentioned two, which is kind of the v3 F in the v3 M, the v3 F I'll start at the top of the spectrum and then work work down the v3 F the cellular one we're we're looking at pricing around this and this. This, again, is the third generation of the sensor. We've pulled it down over 100 times on the cost from what we were originally paying, which is exciting. But yeah, the cellular component there. We're deploying these out largely to residential areas right now in homes of high interest, because we want to make sure that all of the data that we're collecting makes sense in the residential environment. And plus, we're just learning a lot about building the actual building actual cellular detection logic. So that's kind of how we're testing that. The v3 M or the microsensor is just the Wi Fi the Bluetooth and the on network piece. So no cellular we're deploying those, as I mentioned in a closed beta program, to to kind of close folks right not not we did friends and family first, but now we're kind of moving out to some some of our partners for example, have a have a number of our v2 and micro sensors. And they're actually deploying those out. These are existing home security companies, or existing home security customers, I should say, that are getting a sensor and our standalone mobile app, and we're saying hey, set this thing up, you know, assign your phone to your profile. And then let us know what your experience is. Right. So that's kind of what we're doing there. The product team is learning a lot about just just kind of us as a standalone experience, even though we think that the long term solution here is for us to be an Augment to existing home security products, right? You think about you've got your video detection you've got your motion detection, you've got your door sensors, and now you've got RF detection that you can add to that, right. So that's kind of how we see ourselves. So that's the full sensor, the micro sensor. In addition to that, we are we are looking at ways in addition to the rapid response partnership that we can talk about publicly. There's a handful of other conversations going on that we hope to be able to talk about publicly soon. They're talking about direct embedding of our detection hardware into existing home security products. So for example, if I have a touch panel, a control panel on my home security system that has a lot of the equipment on it that we need to do the detection that we're doing. So let's embed directly into that and maybe add a Wi Fi Bluetooth Bluetooth module here or there if we have to, and that's kind of the third one. And then the fourth is just a software only software development kit, where we develop the on network detection logic. We give that to somebody else who has equipment on a Wi Fi network already. And they can just deploy that with you know, as simple as a firmware update. And so the software development kit all the way on the left side of that we're in conversations right now that would probably have a starting to embed that in q3, q4 of this year. Because that's kind of like crawl, walk, run like we want to go there first. And those are kind of you know, think about residential security is kind of the first market that we're going to play with there. And so we're in conversations right now with making sure the the software development kit that we have is going to fit on their hardware that they have right, what additional hardware might they need in order to run this? What are the compute requirements, all of those kinds of things. And we're working on those integrations right now. So hopefully by the end of this year, we'll have some pretty good real kind of what I would call large scale deployments of this from home security integrators that that you would recognize.

Okay, cool. But that's SDK. What about the like v3? Yeah, when

folks, right? Yeah, v3. We're in the final phase of our hardware turn right now and the supply chain is a monster at this moment. So that's slowed us down. We've hit some delays there. We do hope to have functional v3 units by the third quarter, where we will be able to kind of ramp up our beta, again at a higher clip outside of those partners that are working with us already on the SDK. So that's kind of like a q3 and then we have a small number of v3 F's in hand today, our cellular detection logic is still being brought up on the v3 what's all functional on the v2, we just have to shift to the v3 hardware. And that again, should probably land in like q3. But again, those will probably be around sort of small scale. Internal deployments. We have to be really, really cognizant of which partners we engage with on this, because we have limited hardware. And we want to make sure that we pick partners that are highly creative and are ready to move fast with us. We don't want somebody to sign up and say, oh, yeah, this is great. Let's do a beta with all 300 Have your cellular enabled units, and then they just like forget to deploy them for six months, like we just kind of would be sitting you know, twiddling our thumbs. So we are in those active conversations kind of figuring out which partners make the most sense, but there should be some interesting stuff coming again by the end of the year on them. Okay,

sounds good. Sounds good. Also, I'm interested, you said the closed closed environment, closed space close home. So one of the interesting thing that I ran into in the baton was geo fencing. So basically distinguishing the borders of my apartment and someone else's apartment. Can you elaborate on the accuracy of digital fencing? So for example, can I use this system in let's say, apartment building, where there's a lot of highly packed like devices on top of each other and duplexes, stuff like that? Yep.

Great question. We hear that one a lot. The the unfortunate short answer is that it really does depend. But the better more detailed, long answer is. We are really laser focused today on single family homes. single family detached homes for exactly that reason. There's at least 2 million homes out, you know, across the country, that would be pretty easily deployable. So there's a good addressable market there just just amongst single family homes that are separated so we're kind of focusing there. There are though a number of people in our closed beta, myself included a lot of our a lot of our employees included that have these units deployed in apartments. And depending on the, the, the layout of your apartment, it can actually work reasonably well in an apartment setting as well. But it just kind of depends on the construction of the building and a bunch of these other things. We just know it's going to be a little squishier there and so we're not telling people that what your safe apartment building today, but again, today's mobile app, largely focuses on the on network piece anyways, and all of that works really well in an apartment or a townhouse or a condo. We just know that in the future with some of these passive capabilities as they come out. That's going to just get a little bit you know, a little bit squishier. A little bit trickier in an apartment building kind of a setting. on the residential side, we've gone through a couple of different generations of our geo fencing process, at first, that we actually shipped a geo fencing wand that you could walk around the inside of your house for our sensor to get kind of an idea of what the inside look like. And then you did a lap around the outside. And we collected data, we built machine learning models on predicting in versus out, it was a it was a really interesting setup and a really clever thing that we did there. Over time, as we were collecting more and more of those geo fencing runs. The data team looked at because when when when we asked you to sign up, we asked for the square footage of your house. We looked at the square footage of the house and we looked at our geo fencing runs, and we realized they were almost exactly the same for somebody that was in a similar sized house. And so recently, we've realized that we can actually sort of call it auto geo fence or we can we can set a really good perimeter for you without needing you to do that calibration one step. It's an area of extremely active research. I'm not going to say that we're perfect today. But it's it's something that we know that we can we can improve on in the future. And that's kind of where we sit today. The other thing that that's interesting to us about the way we think about geo fencing is the the sensor equipment that we have is relatively inexpensive, especially when you consider it against like a you know a full fledged simply type system or ADT system or something like that, which is multiple $1,000. The fund that we have is gonna be less than 100 bucks on the Microcenter side. And so that's that's the target that we're aiming for. So you could have multiple of those sensors in a home and everything that we have approached today is assuming one sensor in the middle of the house. And at that point, we're kind of setting up a radius right, which is limiting, but in the future, what we're going to be able to do is if you have multiple sensors in the house, we can start triangulating where the signals are coming from. And so in the future right that's that's where the goal is going to be is if you have multiple sensors in the home, you will be able to get much better insight about exactly where those signals are are emanating from. Okay,

cool. You said that the the the circuitry and the sensors itself will cost less than $100 Is it the like the sticker price and MSRP price or should i What will you charge for the for the sensor is like 100 bucks.

We Yeah, we're looking to charge less than $100

Okay, cool sounds good. Also I just lost the thought

while you think about your question, Nikita, I had a question.

Like how do you handle like multiple metaphors because most of the residential routers are like 2.4 gigahertz and five gigahertz? Do I need two sensors to monitor both networks or one sensor can do the job for us? Yeah, the vast majority of the betas that we've deployed to have the dual band capabilities, but they set the same SSID and it's actually working kind of like a VLAN. And so you can monitor both of those with one one sensor if it's the same network internally. If you have two separate networks for like a 2.4 and a five today, you would need two sensors. We've got like lots of active development in the pipeline for how we could change that because we have the hardware on board to sort of jump onto both networks and look around both of those things. But again, the majority of the people that we've seen setup, overlap on the same SSID and you can carry you can cover both networks. If if that's the case.

Yeah, and I was going to ask you about the money. So you said like less 100 $100. And can you elaborate on the state of the company right now and what's the future will be so for example, as far as the money is concerned, we know that it might be easier for like publicly traded companies to get a bigger funding in order to do stuff. I don't know in bigger scandals, skills or charge less rather than private companies. These yourself you're going public and at some point or yeah, what's your binder?

Yeah, great question. So there's we're of course open to any road that presents itself that is, is a creative and valuable where we are today is we're a pre revenue startup. We've raised a good amount of funding to date to cover us to get us to this point. We are going to be able to announce kind of we're in the closing innings of our series, a fundraising round, which will be closing in the next couple of weeks. So we'll be you know, able to publicly talk about that as soon as that's done, but yeah, very close to closing our Series A. We are interested in a couple of different modalities here like the $100 sensor. Interesting, but you know, that's that's just, we don't want to be a sensor company. We want to be a data analytics company that didn't we just had to build a sensor to get us a data set. Right. And so our model long term is for a homeowner to know who's in their house and who's not that feels like that's a recurring kind of benefit to them. And so we're looking at upselling and the companies that we're talking to are thinking about, hey, if we provide a home security package, we could sell this functionality as a $5 a month or $10 a month add on to give homeowners value right and that recurring monthly revenue is what we're interested in maximizing. So we're not trying to make money on the sensors. We're trying to make money on the data and the value that kind of comes comes out of that on the back end. So we're architecting our company, as sort of an arms dealer to the industry. We don't want to you know, cozy up to just one company and become kind of a feature within that company's portfolio. We would much rather enable the industry with this RF detection technology and anybody who wants to build that into their product is able to so that's how we're focusing ourselves today. Like I mentioned, this, this dataset, like as a technologist, when I see a new data set that hasn't ever existed before, there's almost always immense value that follows after that. And this data set has not existed previously and especially not in the home security industry. And so what we want to do is enable everybody else with that new data set. And anybody who wants to benefit from that can do so. And so we envision ourselves collecting some recurring monthly revenue off of, you know, we want to go business to business partner with businesses who want this kind of capability for them. You can chart or whatever you want for it. We'll take whatever clip we need to do a revenue split or whatever that looks like. But again, we're most interested in the in the recurring monthly revenue on that side. And then, of course, sort of on the side, when you have that machine working. You're collecting a bunch of interesting data that other people find valuable. And so like I talked about earlier, the home insurance providers or monitoring centers, all of those folks would potentially have sort of a reason to buy into this as well.

So, thank you. Also interesting question, isn't it the first thing to say that you're the only company in the field that is doing such detection and data analysis because I know there is some companies that do. Let's say motion detection with Wi Fi sensing. For example, this company cognitive, which is doing Wi Fi motion, I'm pretty sure there are some other companies but can you like clarify what what is unique about you buy it?

Yeah, so very good question and the the Wi Fi motion stuff like I know the guys that cognitive really well, they're they're fantastic people. Same with you know, plume and a handful of other folks are doing similar. What their technology does is at its core, it's still a motion sensor. It's just a different way that they detect motion, right? We're detecting something totally different, which is the presence or not presence of a cellular subscriber ID, right. The the piece that we do that's a little bit unique, is on the passive side, not on the Wi Fi side. We're collecting a bunch of that metadata, which is then very valuable. And the the patents that we have and sort of a differentiation that that we get is largely on the cellular side. Like I care a lot about making sure that the cellular product is one that that has legs, not only kind of commercially, which is what our current focus is but also residentially and throughout, you know, cities and that kind of thing. So nobody else is doing cellular detection the way that we are doing. And if they are, if when you think of somebody like you know, like a, I think there's party squasher is one that does this kind of on on Wi Fi. They don't try and ever give you insights about what's actually happening. They kind of just let you set thresholds like too much or not too much. And then when they count, naively, the number of MAC addresses or whatever it is, and it goes over that threshold, they send you a push notification, that's helpful. Like that's one piece of insight and our system can actually do that. Also, like that's one of the notes that we have in Eckleburg, for example, but what we're really trying to get at is that next level that layer two intelligence, which is okay, given that there are more MAC addresses than usual. Is that because they're more iPhones, is that because their Apple Watches is that because there are six people in the house where usually there's four, like that's the level of intelligence that we're that we're getting towards, which nobody else in the industry has demonstrated ability to do.

Great, great. I mean, you're in a great position to to success at this point, I imagine. Are there any questions that I should have asked, but I didn't? From

your question, as as an interviewer, you should always ask that question. Let's see. One of the things that I would be remiss if I didn't talk about was the caliber of the technical team that I'm working with. I've you know, I spent time at NASA Jet Propulsion Lab, like the people that we have brought on to this mission are absolute, absolute studs. We have before we ever hired an engineer, we hired general counsel to make sure that we were always on the legal side of everything that we were doing. She was the Assistant US Attorney for the Northern District of Illinois, came out of Department of Justice had you know, 700 kind of different prosecutors reporting up to her at one point, she's absolutely amazing on the on the legal side, and then on the technical side, one of our co founders, you know, spent a lot of time in the Marine Corps kind of attached to NSA and agencies, signals intelligence background. We have folks from the NASA Jet Propulsion Laboratory, as I mentioned, some folks on our team have led multimillion dollar DARPA projects to push forward the state of artificial intelligence and machine learning and some of the underlying math that is necessary to make that feasible on compute. So that's really incredible folks that have done sort of computer vision and machine learning for path planning for the Mars Rover. A lot of folks from the Department of Defense, especially defense digital service, so kind of the the cyber SWAT team from that side. I've never been on a more capable technical team than this one. And I'm really, really excited that we get to kind of apply ourselves towards this, this problem. So that one, that one is really, really exciting. The last one that I care a lot about is today, the way the industry is set up. The industry is rife with misuse. of public resources, and by public resources. I mean, especially first responders, first responders are so important and so limited, and we need to be smarter about the way that we employ our first responders. What we are hoping to do with this, I talked about that 97% false alarm rate. If I get a call from home security company, you know, 123 and it says, Okay, there's an alarm at this house, and I'm a cop who's heard 18 of those today. I'm probably not going to give it the attention that I did the first time, right, there's that bias that comes in what we're trying to do is empower an industry with context about what's actually happening when that alarm goes off. So the alarm went off. We called the right person. We know exactly who's in the house. We can cancel that 100% of the time like our demo house. We've installed a bunch of different home security systems in there, but ADT was installed when we got there. That ADT alarm has gone off No, no less than 40 or 50 times since we've moved in. And we didn't always know like, our CEO was the one on the account. They would call him and say, Hey, there's an active alarm at your house. What should we do? He pulled out homeware and said, Oh, look, it's an engineer that just showed up. They didn't know the code. You don't need to send the police. We've cured 100% of the false positives that we have had at our demo house by using this technology, which is awesome. In addition to that, we want to arm them with context about what is happening in the situation that they are walking into. So for example, if it's two in the morning, there's a daughter home, and there's an unknown device in the house. And we call a police. We want to armed police with that information. Hey, this is my daughter. She's supposed to be in the house. There's also an 18 T device that's not supposed to be there. If you see this person that makes sense. If you see anybody else that doesn't make sense, they're not a resident, right? We're trying to give people that context, both on the monitoring side and on the on the first responder and law enforcement side. So it's it's been a fantastic company to work for. I'm excited about what we're doing. There's a lot of Greenfield in front of us. We've got a long way to go. But but if we make progress into this field, like there's a lot of value to be covered there. So I'm really excited.

Sounds good. Sounds good. Um, Ilya, do you have any other question because I think I'm good.

No, I think I'm good. I just wanted to ask a couple questions about like size of the company. I know that LinkedIn says that you have 20 employees. Is that a correct number?

That's still pretty close. I think maybe we have 18 full time employees. We have a couple of contractors that we use part time. And then we also of course, employ a couple of like, design firms for expensive hardware and a few other things. But yeah, 1818 full time employees about right.

Okay, and all of them are located in Chicago or you have some remote

majority are in Chicago. We've got a couple of full time remote as well. Okay.

I guess Yeah, that's it.

Can I I'm pretty sure I missed something. Because I had a lot of questions and you obviously gave us all the information can I reach out for you to you in case I came up with additional questions via via email or,

of course, emails. Great. I'll respond. I'll respond as soon as I have time. Okay, cool.

Thank you very much for

taking this fantastic, great meeting both of you. I'm excited to see where this goes. And we'll be in touch.

Yeah, sounds good. Thanks. Thanks.