Interview with Van West of Vocalytics on a Short Backstory on his Current Business

Hello everyone, welcome to this episode of Conscious Design Podcast. I'm your host, Ian Peterman, author of Conscious Design book, and today I'm excited to have Van West, serial entrepreneur and founder and CEO of Vocaletics, here to talk about the entrepreneur journey and of course Vocalytics and the exciting things you're doing there. Welcome to the show.

Awesome, Ian. Thanks so much for having me.

Yeah, I'm really excited about this. So you're a serial entrepreneur, so you've done this journey more than, more than one time. So would love to hear kind of your journey so far in the entrepreneur world and how that's led to Vocalytics and what you're doing now.

Yeah, you know, I like to consider myself a serial problem solver and I happen to use technology to solve a lot of these problems. But yeah, absolutely serial entrepreneur. I've heard that term thrown around plenty of times. Probably goes back, you know, 13, 14-ish years in my career. Had an idea, didn't know what to do with it, so decided to take that plunge for the first time. You know, I've had a couple of my own ventures over the years, have been kind of that employee 1 as a founder as well, and then all the way up to employee 101, scaling to 200 headcount in the first 6 months I was on board. That's actually happened to me twice in the last 10 years. So I've seen various stages of startups and scale-ups and, you know, those growth and inflection points and really understand, especially from a technology perspective, some of the aspects that it takes to get to that product-market fit in order to scale and to do it as leanly as possible. So yeah, just been a part of various stages of ventures in the past and, you know, probably about 6 or so years ago, I joined the startup as the first key hire, kind of their head of product and business development globally. And, you know, in under 2 years, We took computer vision software to large properties, commercial real estate, and facilities, really anywhere that had security cameras. We were one of the first ones to really make them smart with their existing assets. Just plug a box in at the switch level to make sense of those customer journeys and utilization of assets on properties. We even did it for some large national retail accounts as well, from the likes of the Mall of America to Downtown Disney, the Disney Springs, all the way through the Walmarts and T-Mobiles and H&M's of the world. Realized, you know, a ton of opportunity around the visual space that had started to become a little bit more commoditized as we were some of the forefront pioneers doing all that edge compute. Realized that most of these connected cameras, IP-based cameras, also have a microphone, but nobody's making sense of the sound. That's kind of what led me to Vocalytics here.

Yeah, that makes sense. You know, we're very focused on the visual, right? That seems to be Yeah. Be a huge focus. Um, you know, you talked about doing this a couple times, starting with especially with the camera portion of it. Is there anything, as you've gone through this journey a few times, are there any key elements that you've kind of taken with you that, that do cross those different stages? Because you've worked, you know, employee number one, been a founder, and, and later. Is there anything through that process you've learned that that kind of guide— there's, there's a constant between it, or is it just all across the board there's— there is no constant, it is just other than change?

Yeah, you know, the, the only constants that I've had kind of as that guiding compass throughout each step of the journey, um, personally, and then for each of those journeys I've been another chapter in their books for, has really been to lead with purpose. And, you know, I mean that in a few different ways, but, you know, Simon Sinek, one of his first books, Start with Why, really hit me as that concept of starting with that purpose, really understanding the why you're building something or working on a project or a task or an opportunity, whatever it is, starting with that purpose and then leading out to the, you know, okay, so what are we going to do? And then how are we going to go about it? The strategy and the tactics aspects, whether I'm building out a marketing campaign, a sales organization, a business, frankly, with Vocalytics as a startup, really leading with that purpose of why are we solving these particular problems.

Right. Yeah, why— the why is, is, uh, very important to be, to be successful. And it's, it's something that— and when I'm talking with, with people that are starting up companies now and even existing companies, it's, it's such a huge topic because that's what customers more than ever are now aligning themselves with their why with the why of the brands that they're using and, and looking at, looking at that. So even from a, you know, marketing standpoint, if you don't have a really good why that's believable, right, you can't just, you can't just make up one after the fact and it doesn't work as well, that it's really hard to be connecting to customers as well and be able to actually grow the business. Now with Vocalytics, you've, you're, you founded it, right? You're from ground zero working through it. What's, what are the factors that you've brought into your why that's been able to allow you to grow? What's your, what's the why behind it that you're, that's making everything work?

Yeah, great, great question. So, you know, I'm gonna say there's a few different elements to unpack here. And when it comes down to startups and the journey overall, there's kind of 2 things, 2 factors that I've talked to a number of VCs about. And, you know, the 2 leading factors for, frankly, any venture investment come down to team and timing. And the timing really comes down to the market conditions. So, you know, when I founded the company, we had some inbound LOIs from the likes of Walgreens and the Walgreens of the world. And, you know, we were really doubling down into the retail sector, frankly, having a lot of experience selling camera software into retailers, at national retailers at that scale, had a lot of those relationships to pull from to experiment on audio concepts. And, you know, frankly, it led me down a path of, can it— and really, can was the first leading question— but is it possible to bring call center AI to those frontline service level interactions, whether it's the front desk at a Hotel or resorts, understanding that conversation and adding color and context to it from a discovery aspect. Are they recommending the steakhouse that's on the property or the other amenities? And frankly, what are the guests actually requesting or asking for at that point of service? And same thing, you know, at the returns counter at a Target or Walmart or at the point of sale, you know, is there an upsell opportunity or a service recovery aspect of that conversation? And then of course, right when we founded the company, the pandemic hit. So all these—

That timing element.

Yeah. And so the market conditions led us down another path of as much as we were filtering out that background noise to get the clean conversation of that, you know, in-person interaction, we actually decided let's shelve that for the interim and, you know, focus less on the words and speech and conversation and voice, Let's start understanding all the nuances that occur in and around and between those words. So less about the voice and the sentiment interaction and inflection and tonality and all of that that can be extracted from voice, let alone the context around what words are spoken. But let's focus on all those background noises since retail and the world of— frankly, the world has been more socially distanced than ever. And when you're in person, you're wearing a mask now anyways, still after 2 years. So, you know, let's, let's focus on the background sounds and noises. And it really led us down a path of healthcare that's been so incredibly relevant, not just for us to be helping a lot of hospitals to understand the background noises in noisy patient rooms to improve the patient experience, frankly, accelerate recovery times, improve outcomes. And by improving the patient experience, that's actually improving a lot of the Medicare reimbursement, happens to be a third of all hospital revenue. So a large factor in that survey known as HCAHPS to measure the patient experience in hospitals, what we've understood is an 8th of that, one of the 8 domains boils down to quietness. And so why quietness, we asked? Well, it comes down to restful sleep, faster recovery times, improved outcomes, therefore better patient experience, therefore higher reimbursements. The hospital rooms are really noisy places that they've never been able to solve for what's creating the noise in those spaces. Doctors and nurses have such alarm fatigue that they don't even recognize a hospital pole has been beeping for 45 minutes at a time. So we're the ones that are making sense of that audio data in those patient room environments to improve that care delivery aspect.

Right. Yeah, that's something that, uh, when we first start— first started talking was was something that really caught my attention because it's especially in today's— you go into downtown anywhere, you don't— there's— and then you walk into the middle of nowhere and you can suddenly, if you were to actually contrast those rather than gradually traveling between, you'd realize just how much noise exists. And it's right, something's making the noise. It just doesn't appear out of nowhere. It's being created.

Yeah.

And you guys are the first that I've ever heard of thinking about, well, how do we start looking at that, what we as humans really just begin to treat or already treat as just background noise and pulling that information out of it? It's another dimension to our space that, you know, we talk about noise pollution and we talk about, you know, noise reduction and things like that. But looking at it as a dataset to really start to pull information and understand, you know, why is the noise there? What's causing it? What is happening? What's happening around it? And that kind of data, I think, is just something that— well, we have a lot of raw data. And if we were to use it, there's a huge amount of opportunity. And, you know, I love that you guys are in the medical space, and it's— I wouldn't have known, but It makes perfect sense if you're in a hospital working as a nurse, 8, 10-hour shifts. I don't know how long the shifts sometimes get. You would eventually start ignoring the alarms. You would, because it just, if you have tons of rooms, there's just noise, right? And so it's great that you guys are being able to kind of sort out, because like you started, you started down the path of, well, let's do voice recognition. Let's figure out the voice portion. And that clear audio, which I think everybody's super focused on, you know, Alexa and things like that. Everyone wants clearer vocal recognition. So how are you started down this path and you got into this? So where do you see this going? Are there different areas? Obviously healthcare is a huge, huge section that needs help, but are you start— where are you exploring and seeing other fields that You know, this, this data is there and just not— it's kind of, it kind of feels like an untapped resource that we're, that we're ignoring right now.

Yeah, no, 1,000% is, a million percent is an untapped source of valuable information. Um, noise, if you just unpack that word a little bit, it's commonly understood as a distortion of sounds. And what we're doing is cleaning up that noise to understand and classify a large library. When I say large, I mean more than 5 100 ambient acoustic sounds that occur in any space. And the possibilities really are endless for us. That's a lot. We started in healthcare. It's a large library we built up. And we do it really accurately because we give away the entire library as a part of every single deployment. It's not like we partition it out and say, you want a beeping hospital pole? You want a siren? You want glass breaking? You want aggression? You want each one of these individual sound profiles or buckets of them? No, no, no, we, we don't do that. And frankly, even for security use cases, safety as well, you think about, you know, if we're looking for gunshots in the waveform, in sound, if there's an active shooter situation, well, if we're not actively looking for a car backfiring, we're going to have a very high degree of false positives. So, you know, even in a patient's room—

Yeah.

In a hospital room, just as another example, coughing, wheezing, heavy breathing beyond a certain decibel level over a certain period, I'm going to say over 30 seconds, could lead to classification as an upper respiratory distress event that can now be monitored, tracked, and actually enhance that care delivery process for physicians, clinicians, and ultimately caregivers at the end of the day. So there's lots of different ways, you know, don't think of this as like if this then that, because you don't want an alarm or alert to go off every single time someone coughs or sneezes.

Right, you're trying to help, you're trying to help fix that problem already.

Right, exactly, exactly. So we can define those thresholds with our end customer. Frankly, they can define them for us, and then we produce a real-time alert. And, you know, you throw it out there for your, uh, for your listeners that are actually watching the video, and we got a little case of AirPods. And even though we're a software company, we're an enterprise SaaS company, we built a little widget to say, Mr. and Mrs. Customer Hospital, all we got to do is plug this little guy in, operates over cellular, processes everything on the device. That's, you know, one other topic I wanted to touch on too is in a high-trust environment, we have to be dead set and focused on privacy. And for us, we believe that's a fundamental human right. So we do all the processing on the device. We're not like another Alexa service that pipes everything up to the cloud and figures out what to do with it after the fact. We're actually doing all that intelligence on that lightweight little device that I just showed you. So You know, not even to say that we have to use our device. We can integrate with a call button that's already at the bedside or a security camera that has a microphone embedded. It's really our vision to unlock the value in everyday sounds from any connected microphone sensor.

So you're, so you're gonna, you're able to connect to existing hardware if someone has, you know, a camera with the mic built in, which I feel these days most smart home Cameras at this point have, have mics built in. It just kind of seems to be paired, paired together a lot in different technology. So you're, you're able to access that existing network of hardware and then be able to process it locally, which is great.

Yep.

As we talked about, like privacy, data privacy is huge. It's for individuals, And there is this movement of, it should be a human right. It is yours. You should keep it. And from a business side as well, I mean, thinking about a healthcare application, that's all patient data, right? That's all potentially medical information that you're handling. I love that you guys are looking at this unique solution of let's have it on site. It's not, you know, I, I think, I think of cloud as a tool and I think we're all a little— we all think the cloud is the solution for everything. And I think it's kind of like every other technology, right? We, when we get out a new shiny thing, we all go, well, let's make it everything. I'm talking with people about in the AR and VR space, the same thing is happening. People are trying to apply it to everything. And it's just not, you know, it's a tool. And so it's nice that you guys are thinking of this from, you know, the ground level of what does our customer actually want? Like, what are we really doing with this? And you guys have come up with the answer of, well, on-site is actually best for this application, what we're trying to do with this. As you're going through this, you mentioned being able to— you don't just check for one thing, you check for a whole library and you're combining that. So is that something that is just going to grow over time and just slowly your library grows? And do you apply that to every customer? Do you have that or do you kind of individualize their profiles in some way?

Yeah, really good question. And, you know, when we're having the same conversation in 5 years, I'll be able to have a more definitive answer. We're about 2 years into the business and, you know, about a year into doing the actual deployments with end customers. And so I wouldn't say that we're in a position to say this library is going to grow to 10,000 sounds and it's going to have to process 10,000 sounds in every deployment. I can see that library becoming more specific around use cases, or more specifically around certain environments. So, you know, the acoustic cleanup that really goes in, that's a lot of our intellectual property that goes into, you know, you think about artificial intelligence or machine learning, quality in, quality out, as far as the input to the output.

Yeah.

Garbage in, garbage out. So if you just get a bunch of noise, we're not gonna be able to make sense of it unless we can clean up that space or that environment. Beautiful thing with sound is it exists Everywhere, so that the possibilities are limitless for us. We just need to think very specifically around a patient room as an example. May not have as frequent of an active shooter environment, but in a VA facility, think about safety.

Mm-hmm.

Think about crying, screaming, yelling for aggression detection. When a vet may have PTSD, it would be very helpful information for the nurse that's going to do their rounds and, and check on that particular vet. If they understood that this vet is trending more on the aggressive side, to walk into that room and engage with that patient in a more meaningful manner. Same thing with, you know, grandma drops a coffee mug in a senior living center, and all of a sudden there's glass breaking that was detected. Well, rather than walking into the room and there's a call button that got the nurse there, or, you know, an alert went out to the reception and there's a caregiver that's showing up asking, hey, grandma, what's going on? especially in a memory care unit where she may not even remember what happened.

Right.

Then it really just comes down to we're providing that intelligence for faster response, better situational awareness, and overall total resolution of the actual incident as it's occurring. And then our customers are also looking at that as you don't know what you don't know. These sounds have never been monitored in these spaces before. So what's the trending data look like? You know, how frequently are doors opened and closed, or there's a squeaky wheel on a hospital cart? You know, I'm giving a lot of great, healthcare-specific examples because they're really noisy spaces, but they apply anywhere as well.

Yeah, that is, that's something interesting to think about, right? We've never, you know, nobody's sat in a hospital hallway counting squeaky wheels, right? Nobody's, nobody's actually sat down and done that. And it's such an untouched space. It seems like there's a lot, you're kind of walking in Blind, right? You're in there. We don't have a data set to say, well, this is the average number of X because we, we just don't track it. It's all to us. It's, it's that background noise. So are you in this process, are you finding that there's a lot of exploration and testing to create these kinds of baselines to even build a model? Is it, is it a big in-depth process, or are you able to— How do you go about figuring out what normal, right, is when we all ignore half of that?

Yeah, and I think you touched on a really good piece earlier around the cloud and, you know, what is the cloud? The cloud is, you know, an elastic set of tools and, you know, really the connection to the internet that's been finicky a little bit today here at the office. It's really just the pipes. at the end of the day. So, you know, I just look at the pipes as, is the road actually necessary to produce intelligence or can it be done within the vehicle that you're in, even stationary? And so for us, you know, a big reason why we have our own widget is to really expedite the procurement process for hospitals that, you know, to tell them that, hey, we've got a little widget that's much like an air quality sensor for more than 500 sounds in any one of your spaces. Suddenly we get, we get kind of placed into that environmental sensor bucket rather than a medical device. So less about the FDA concerns or, you know, high trust from that perspective. And the fact that we're able to demonstrate that we do all of this even offline, we don't even have to have access to any kind of connectivity whatsoever in order to produce all the intelligence on that tiny little device or just local in the facility. The fact that we can demonstrate that gives them an ultimate level of trust Knowing that all the data is safe and it's processed and all the intelligence occurs locally. It's not piped off to some cloud like an Alexa.

Yeah.

So, so that's— those are kind of some of the bigger pieces. And then we actually give the nursing station access to replay those sound moments. So don't think about our service as a way to dial in remote and listen into that environment. We're producing that intelligence and then storing it as just a sound clip, just that little sound bite, whether it's coughing or wheezing or heavy breathing or the sounds grouped together as an upper respiratory distress event or a beeping hospital pole, we give our users, frankly, the nurses at the nursing station, the ability to play back those sound moments and validate or invalidate. So all the training and that kind of human-in-the-loop element is actually our customers, not even us, that helps to build a better product for them, helps to improve the accuracy and detection in our models. And it also helps to, you know, service our clients better when we understand how they're interacting with our solution and how valid or how validating or invalidating they are on our hypotheses. And these are the trending sounds in your facility. So, you know, there's, there's those aspects, but the 3 sounds that we classify— we talked a little bit about voice and speech and words and all of that and conversations and interactions in the very beginning. But, you know, what we've chosen to do is we can classify male, female, and child speech. We don't even give the user the ability to play it back. We actually redact all of that audio. So if there are words, you could never actually replay them as well. So we actually redact those from the waveform. So we'll tag it as male, female, or child speech. We know it's a sound in a space at a time. You know, there's a prediction score, there's a decibel level, and there's all sorts of telemetry data around looking at that spectrogram and analyzing it and classifying it that we do grab, but we don't ever look at the words. You know, you can listen to a cough, or you can listen to heavy breathing or wheezing or a squeaky wheel on a hospital cart or glass breaking or a gunshot and really validate or invalidate it with that human ear. But if it is labeled as male, female, child speech, it can never actually be recreated. That actual clip is redacted from the entire service. So that never hits a cloud, that never goes anywhere. So, you know, from a privacy perspective, it's a sound in a space at a time. There's no speech involved. Right.

Well, and with you never leaving, never trying to go to the cloud, means that there's never a data set in transit, right? There's no— it doesn't leave the building, so to speak, um, which I think is a prob— probably what you said is a huge factor. And, and being able to run You know, hospitals, from what I know of them, is, you know, they're, they're designed to withstand and remain operational should the rest of the city grid go down, right? They're designed to keep things running. They have backup generators. There's a lot of work that's put into ensuring that the system can run as long as, as humanly possible should there be external events. to shut down access. So seeing that this is, you know, not cloud-based, right? This is a great one of those cases where you need to have that ability because otherwise, you know, like having internet issues. Well, if the internet goes down, do you want this tool to not work anymore? And that seems rather important for it to keep working. Even if Wi-Fi goes down, we gotta have this. The system's still in operation somehow, which is great, and it's a great application for, for on-site, I think. So as, as you've gone through this process, is there, you know, anybody that's looking at this thinking, you know, how, how can I use this data? Is there any limitations to this that you, that you are aware of right now that You know, maybe certain scenarios where it doesn't make sense. Is there such a thing as too much noise in a space for you to be able to parse out, or kind of where are you seeing the edges of use, if any?

Yeah, great question. I'll unpack it real briefly in 2 different ways. You know, number one, if you've ever been to a concert hall or watched a live performance or been to a sporting event like the Super Bowl we've got coming up on Sunday, Yeah, we— Think about really noisy, really noisy spaces like that where if you've got an artist performing on stage, especially at the halftime show, you know, if you've got a microphone that is near the stage, it's only going to hear that performance. It's not going to hear, you know, glass breaking or potentially even a gunshot that occurs in the back of the room, or especially coughing or sneezing or anything like that. That's oh so relevant in a COVID-ridden world. So, right, you know, when you, when you think about the limitations of the technology, it comes down to that sound space. What is the environment? And if it's a really noisy environment, we have ways of strategically placing these connected microphone devices, more of them within a noisy space, so we can understand the nuances on such a microscopic level for that entire space, just because it happens to be noisier. And then the other limitation that I will say is that our little widget that we built, it's a commercially off-the-shelf available board that's kind of a variant of a Raspberry Pi. So it's not proprietary to us on a hardware stance. We load our software on it, and the software is really where the magic happens.

Right.

But, you know, we found this board, 3D print a case around it, makes it really simple to get off the ground and do it very quickly. The little board we found actually has 4 different microphones on it. So if you're just using, relying on like a security camera that has a single microphone on it, or the call button that's at the bedside in a hospital room, the advantages you get from like an Amazon Alexa omnidirectional mic array, just several different mics spread out in an array, you don't just get the noise reduction, but you actually get time of flight as well as direction of arrival data. So think about if there's a gunshot and a security camera hears a gunshot, we can probably classify the type of weapon that was used, whether it was an assault rifle or a pistol or a machine gun or, or whatnot, just based on that sound signature. But we won't be able to give you that accurate location data on where this came from. Sure, this camera over here heard it, but what direction did it come from and how far away? A single microphone won't tell you that. Even based on decibel levels, it still won't be as accurate as having multiple microphones in an array like the little widget that I showed you. Essentially, on this little guy, we got 4 little microphones. So think about like north, south, east, and west. We get that direction of arrival and time of flight, so we can actually sense what direction it came from accurately, as well as the approximate distance. And then given the indication that it was a gunshot, here's the actual weapon classification. So, you know, if it could be a handgun that could be concealed, or if it's a different level of threat like an assault rifle. So there's some limitations in, you know, what hardware is required to extract some additional feature sets from the audio. And, you know, you think about the sound space, if it's a really noisy space, especially a large space like you know, a concert hall or a music performance or anything like that, if you put a microphone next to a speaker, all it's going to hear is what's coming out of the speaker. It's not going to hear anything that happens in the back of the room.

Right, right. There is a point where the sound wave just is maxed out and there's— there is no variation with what you're picking up anymore. Um, perfect. Well, this, this is an amazing technology. I think you're— you guys are— you seem to be alone in this space, which is great for you guys. I'm sure more people will pick up on it though, because of how valuable it is. It really seems like such an untapped resource. So, you know, for anybody that's listening, watching, who wants to explore how your guys' system can give them data, because I'm sure there's some exploration of, you know, what can be done, what can be brought together into a smart dataset.

Where would—

where are people best to find you, contact you?

Yeah, I'm gonna say easiest way is our websites. Feel free to email me directly as well. It's the same domain. It's just van, V as in Victor, A-N, that's me, at Vocalytics, V-O-C-A-L-Y-T-I-C-S, so vocalytics.ai. So vocalytics.ai is our website. van@vocalytics.ai is my personal email. Feel free to reach out to me also on LinkedIn. You know, look me up. It's just Van J. West. If you follow the LinkedIn or just find me on Van West.

Perfect. Well, we'll make sure those links are in the description too, so we can— everyone can just click those if, if that's easier. And yeah, it's been great talking to you, learning about what you guys are doing and, and the capabilities of this technology. And yeah, I'm excited to see where you guys can and take this further into using more data, getting more data out of the sound that we all ignore for the most part.