Ep. 13 - Industrial Safety AI: Stopping Accidents Before They Happen

Safety is usually treated as a lagging indicator where leaders only react after a claim is filed or an injury occurs. This reactive stance is a massive profit leak and an avoidable human cost that traditional safety walks can no longer manage in complex industrial environments. Vernon O'Donnell, CEO of Voxel, explains how computer vision and AI are flipping this script by identifying risks before the collision happens.

We sit down to discuss the practical application of visual AI in distribution centers and manufacturing plants across the globe. The conversation covers the reduction of workers' comp claims through ergonomic heat maps, the removal of "noise" from thousands of daily alerts, and the transition toward 90% true automation in safety monitoring. Vernon shares how Voxel identifies operational choke points—like poorly managed dock doors or incorrect forklift governors—to solve the root causes of near-misses.

The unglamorous truth of industrial technology is that many companies fear a multi-year IT nightmare that never delivers. This episode highlights a shift toward lightweight edge devices and software integrations that provide measurable value in days rather than months. You will walk away with a clear understanding of how to treat safety as a strategic lever that increases throughput and reduces employee attrition rather than just a cost center or a compliance checkbox.

Transcript

Meet Voxel CEO Vernon ODonnell

Preventing Accidents Before They Happen

SPEAKER_00 0:08

I want to clarify that this podcast is distinct from my responsibilities as a professor in the Sam M. Walton College of Business. Nonetheless, it aligns with my aspiration to provide practical insights to professionals and business by showcasing companies and people that can enhance your ability to manage, lead, and strategize and market effectively and the retail value chain. And now without further ado, let's get into the exciting episode. I have with me today Vernon O'Donnell, who's the CEO at Voxel. He has extensive executive experience with B2P SaaS companies. And uh Vernon, thank you so much uh for joining us today. Really appreciate it. I appreciate you having me, Matt. Thanks for uh thanks for bringing me on. Yeah, it's really impressive um how Voxel is growing so quickly. Um I know you all have raised about 60 million in um funding. Uh you've got uh about 250 sites deployed worldwide um and a 65% average reduction in uh workers' comp claims. Um those are all very impressive numbers. And uh one thing I'm curious about is you know, so I mean you have a platform that's using uh visual technology along with AI to find safety issues. And so what I'm wondering is what does Voxel or what problems does Voxel solve that most companies don't see until after the accident?

SPEAKER_02 1:55

Yeah, great question, Matt. So I actually want to maybe reframe it. So if you think about safety, it's a priority for all of our customers and really for most industrial businesses out there, but it's also an outcome, right? They're core businesses and distribution of goods, manufacturing of goods, and really everything in between. And safety has happens as a result from all of the unseen challenges you face in those environments. Human problems as it relates to how the body is moving, right? Ergonomics, or are they wearing their proper protective gear? Uh, vehicular problems are people speeding on the forklifts, uh, are people, you know, are there choke points where they're having near misses or collisions, and then environmental issues, spells on the floor, doors left open. And all of those are outcomes of businesses doing their core operations. And today, safety is someone has to walk the floor, look for problems, people are cognizant of that, they change behaviors, or an injury happens or a claim happens, and you do a post-mortem and try to understand what occurred. The beauty of using a computer vision AI solution like Voxhole is we we invert that. We take it to real time, we get to root cause, and we help identify problem areas before something happens, right? So that example of a choke point where two forklifts are crossing, they're having near misses over and over again. Um, before that collision happens, before they have damaged goods, before somebody gets hurt, we're telling them that they have that choke point. And the really cool thing is not only are we seeing that from a visual AI perspective, we've actually trained our models to then provide recommendations on how do they fix it, right? You know, more clearly demarcated lanes on the floor, widened aisles, more visible stop signs. Uh it's that's the beauty of really thinking about safety as a strategic lever versus just safety as an outcome.

SPEAKER_00 3:41

If Voxel didn't exist, what are companies doing instead? And where does that break down?

SPEAKER_02 3:49

Yeah, so what companies are doing instead are a couple of things, right? They have certified safety professionals or safety experts who are designing programs based off of best practices and systems that have been um built in a lab, and they're trying to apply those to those physical environments. They're also investing in um physical uh changes and tools, right? Better safety equipment, better harnesses, um, things along those lines. But again, that's all outcome-oriented, right? You're investing in the outcome of the problem, not the core solution of the problem. And so today, people put a lot of money, effort, energy, focus on it, uh, but they're still inherently constrained by only addressing problems after they occur, only understanding where risks uh exist when someone's already hurt, right? That that's the beauty of this whole revolution of AI. And I think we can take a very human-centered approach to AI. It's not about replacing human workers, it's not about uh automation in the warehouse. It's about truly providing intelligence, right? And using that intelligence to change how workflows happen. Uh let me let me digress to an example. We had a customer, um, a large retail customer in the grocery space, Fortune 100 company. Um, and they saw a significant number of near misses at their dock. So forklifts and pedestrians interacting. They knew that was a problem. They'd had some collisions, they'd had some serious injuries, they'd had some major, major claims, but they couldn't understand why that was occurring, right? And they can safety experts can come in and provide all the coaching and all the counseling and all the corrective action they can, but it just wasn't changing it. Um, they turned voxel on, and ultimately what we determined was that this was a safety problem, but it was a safety problem tied to an operational challenge. They had a load-unload process at the dock doors that was leading to cluttered floors, improperly stacked pallets. It was blocking lanes for the forklifts. They had no other way to traverse except for veering close to the pedestrian workers, the on-foot workers. And so it's a great example of like the AI helped them actually solve an operational problem, which was downstream having a safety issue. That's not really going to be achievable with the approaches that people have today until something occurs that that leads them to kind of like take a step back and really reassess their business holistically and in a very costly fashion.

SPEAKER_00 6:15

So, so Vernon, what happens between actual camera footage and then action on the floor?

SPEAKER_02 6:22

Yeah, so great question. So the way Voxel works, we're purely a software um uh solution. So we integrate with the existing cameras. The great thing is we're very flexible in that regard. We have some customers, we don't encourage this, but we have some customers that have analog cameras where they're sending analog video or digitizing and running our models. Please don't do that. Uh but for the most part, we integrate with our existing cameras. Uh, we do a little bit of processing there in the facility. Um, we add some security protocols, we do some limited uh versions of the algorithms, and then ultimately we stream it back to our proprietary AI solution of the cloud. Um, and in a matter of seconds, about 15 to 20 seconds, um, all that video is analyzed in in near real time on a frame-by-frame basis. And we're returning back clips to the customers. And really, there's two modalities for that. For things like ergonomics issues, you don't really need an alert every time somebody's improperly bending or every time someone's lifting overhead. What you need there is data, right? And so inside of our application, we're quantifying that, we're pointing them to heat maps, we're pointing them to particular shifts or doctors or cargo types that are correlated with ergonomics and providing recommendations. So there's a web application with a really robust reporting ecosystem. But then there's other types of incidents and risks where they need really quick interactions. Like someone's working at heights and they're not properly harnessed in. In that scenario, we're going to send an SMS, we're going to send an email, we even have IoT integrations, potentially flashing light in the warehouse, letting them know that there's a risk that needs to be addressed because there's catastrophic risk at play. So it's really about both of those things, providing really, really clear data with actionable insights that allows them to make systematic changes to drive down things like ergonomic risk or vehicular risk. But at the same time, if someone's at risk of a catastrophic, right? Falling, getting run over, they need to know much faster. They need to know across a variety of different delivery mechanisms.

SPEAKER_00 8:22

So, how do you keep from overwhelming people with alerts?

Making AI Work In Messy Warehouses

SPEAKER_02 8:28

Yeah, it's a one of the first questions we get, because uh one of the cool things about the tech when we turn it on, um, we've got over five billion hours of video footage at this point we've analyzed. And so I mentioned our algorithms are proprietary at Ambleton House. We brought in a little bit of the open source, but we turn it on and within 24 hours, we're seeing risk. And we're seeing risk in that specific site. And it just gets richer and richer and more uh more tailored as it goes because the AI is built for learning at the specific site level. It's not just broad application. Um, but the the other thing there on the overwhelming thing is you'll see tons of incidents right out of the gates. Uh people's environments are never as safe as they think they are. And that goes back to that data piece, right? For ergonomics, it's not about sending them every clip. It's about how do we consolidate that into you have a problem in the second shift uh at the fresh dock where people are overreaching. We tell them that, we point to it, we point to a specific system problem, we provide one or two representative examples. The AI will actually select that and provide them one or two clips because video reinforcing the data is a really great way for humans to digest information. And then we provide recommendations. And so, really, we're saying don't look at the 3,000 examples of overreach. Here's two representative ones, here is where your hot, your, your hotspot is for this occurring, and here's how to solve that. So that's how you really cut that noise down and make it much more actionable. Again, independent of things that are serious where you want the alert. And so we always counsel people, you know, the great thing is there's a lot of operators who take safety super seriously, and they're like, we want to see everything. Let us let us look at every incident. Like, you're gonna get 10,000 alerts. They're like, no, no, our environments are safer than that. Like, they're not, trust us. You have a million eight square feet, you're gonna get 10,000 alerts in a day because there's just a bunch of stuff happening. And so we also then provide risk scores as well, is another thing I forgot to mention, you know, severe risk, uh, high risk, medium risk, low risk. And so that way they can also filter by that for how they interact and digest the information.

SPEAKER_00 10:32

I had been hearing about computer vision and um, you know, logistics settings for a long time. So it's it's really cool to see that you all have made it uh effective. But I also know there's lots of messiness in these kinds of environments with uh speed of uh forklifts and people and uh uh lighting, uh congestion. Um it seems to me that would might make it less reliable, those kinds of factors. How do you how have you dealt with that uh in the past?

SPEAKER_02 11:10

Yeah, I think there's there's two things at play there, Matt. And and you're right, these industrial environments are very messy. Um just the advancement in technology in the last two to three years and the the depth and quality of what's available from an algorithmic perspective, from a compute power perspective, um, it's been a sea change. Uh and so by way of my background, I actually started in computer vision AI 11 years ago. So I've been in this space for a while. I was actually in sports technology, though. So working with a variety of professional soccer and basketball leagues around the world on athlete performance. And what was capable a decade ago versus what is capable now is it they're not they're not even comparable products. Like so part of it is just a massive technical advancement. But the other piece, and it I alluded to this previously, it goes back to the depth of understanding, right? Where we separate ourselves is as I mentioned, we have over five billion hours of footage. Um, we've been training these models for years, and there's an incredible amount of nuance being enlightened hundreds of facilities and thousands of cameras that allows you to start to parse out the noise from a technical perspective. Uh, so the cool thing is the AI does a self-configuration. And then we do a little bit of human expertise uh as well. One thing we're proud of is we have a team of certified safety professionals that are full-time voxel employees. Uh, we have full-time former operators, warehouse managers, and the like who actually work for us. And they provide some of that expertise on top of the AI to say, yeah, this looks right, or oh, you're missing this piece of context here. And the beautiful thing is, as they do that, it's then trained into the model for the next site, the next site, and the next site. So it's really about punching through the messiness with technical improvements, depth of information, historical training of the models, and then true domain expertise. Uh, and the last thing I'd say is people get really excited about all of the possibilities of AI, and they are nearly limitless in terms of the mindset that you could apply to it. But the best products are ones that are very much purpose-built, right? You understand the space, you build the technology around that space, you go really deep into that space. Because having been someone who started my career as an operator at Granger, is where I started my career out of grad school, um, you understand that you don't have time to be the person to decipher the information. Go back to your noisiness question. And so if you're an AI solution that's like, oh, we'll build whatever you want. I don't want to have to tell you what to build. Yeah, I want to buy the product and have you tell me what is going to happen because by the way, my job is something else entirely. And I think a lot of young companies make that mistake and they fall in love with what's possible instead of falling in love with how to solve an achievable problem using technology.

SPEAKER_00 13:58

You know, one really popular concept for some time now has been this idea of uh automation with human in the loop. Um and what's the balance there uh with Voxel?

SPEAKER_02 14:13

Uh great question. So, human in the loop, I mentioned my sports background. Back in the day when we were working on this a decade plus ago, you'd have hundreds of humans in the loop, right? Hundreds, hundreds and hundreds, and almost every clip would be uh human-reviewed at some point. I'm happy to say here at Voxel, in the last two years, uh, we are now over 85% true automation, meaning there's no human ever looking at the clip before it goes in front of a customer. Um, and the great thing about that is we've sacrificed nothing from a precision perspective, uh, a recall perspective, the notions of accuracy, if you will, um in doing so. So we're achieving 97% precision plus with end-to-end AI. Um, one of the things you we've done, and I don't want to go overly into our secret sauce, uh, we actually built a middle algorithm that essentially learned from human in the loop and approximates a human in the loop on top of the AI. So essentially we we built AI to do human in the loop review. Um, and so it's really cool because that first 15, 20 seconds is the initial analysis from the algorithm. And then the AI does another review on top of that, approximating human in the loop aspects. That's about another 20 to 30 seconds, it refines it even further. And so by the time it actually is in the product, it's that 97% plus precision. And like I mentioned, over 80% of our stuff is never even looked at by a human. Um, and the goal by the end of this year is over 90%.

Fast Implementation And Time To Value

SPEAKER_00 15:38

That is impressive. Um talk a little bit about um implementation and how long it takes to see value.

SPEAKER_02 15:47

That's a big thing um for us to focus on. I think it's a big thing for technology companies in general in this new era uh of AI deployment. So we are um typically live, but it's a small edge device, it plugs in. There's some work with IT to go back and forth. It's called two to three hours of phone calls, excuse exchange of emails, um, then it grade to their camera system. And then once we're integrated, uh day zero, the algorithm provides results. We turn it on internally at 48 hours to do an internal check, and then it's customer facing in five to six days. So that whole process for a site from end to end takes a little less than three weeks. And that includes back and forth on the IT calls, uh training on the product, which is very intuitive. It's very lightweight, it's one 30-minute call. It's it's very explanatory, but it's fast. Uh, we have one customer that's in over, we're at over 50 sites in 16 countries. Um, and we did all of that in about four and a half months. Um, so it's it's it it really rips. Implementation challenges of the past, like we always tell people at these big companies, we have a lot of major Fortune 100, 200 customers, and they're like, you know, we got to plan for our IT resources, and it's gonna be a two-year deployment. Like, you know, and no offense are WMS customers or companies out there, but this is not the legacy WMS warehouse management system or TMS transportation management system implementations that take months, if not years. It's days and weeks at most.

SPEAKER_00 17:13

You know, detection is one thing, behavioral change is another. And I know that, you know, real behavioral change oftentimes, I mean, it's really driven by change management. Um, so it has to do with the leadership in the facility and so forth. But how does Voxel help with driving that kind of behavioral change?

Before And After Safety Results

SPEAKER_02 17:35

Yeah, it's a really, it's a it's a really salient point, a really important point. So we do a couple of things. One, um, our application provides recommendations on what to do about problems. So it's not just seeing it, it's recommending what people need to do. Because so often, operators that aren't safety experts, they have the right intention, but they don't have the right know-how. And so helping them know what to do is a big component of behavioral change. We also build collaboration flows into the product that are both external and internal. So our safety experts can provide feedback, coaching, um, guides, examples, best practices. Same thing for their internal users. And so we're providing guided recommendations, we're providing collaboration paths. Uh, and then there's also validation of the beautiful thing with computer vision. If we're recommending that the stop sign is more high visibility and more in a different placement, we can verify with the AI whether that happened. So it creates an accountability loop. So it's really about stick and carrot. Uh, you know, carrot. Oh, the other piece I forgot to mention is we do also identify positive behaviors. Um, and so, you know, an example of a two-man lift where someone's lifting heavy boxes, two people, they're waiting for a coworker to come get it done versus just doing a, you know, a heave from the floor to the conveyor belt. Uh, we'll recognize those sorts of positive behavioral traits as well. And so the combination of providing recommendations, coaching from a care perspective, access and collab to collaboration tools, providing examples of best practices, and at the same time, the accountability of like, did people actually in that change really sees a lot of quick acting behavioral change. And so often it goes back to they just don't know the problem until we point to it. So that's a big piece of what we do today, Matt, is focused much less on here's our whizbang cool tech, and much more on how do you get value from it and how do we see systemic change in your environments?

SPEAKER_00 19:29

Well, your technology uh has been deployed globally and in over 250 deployments. I'm curious, um, could you give us uh just an example of a before and after, maybe?

SPEAKER_02 19:46

Yeah, I'll give I'll give you two. Um, actually. So one, uh, there's a customer with a high propensity for ergonomics claims. They have they have a lot of ergonomics claims to the tune of hundreds of thousands of dollars in one single facility. Um, so the before and after for that is, you know, they're providing coaching, they're showing people best practices for how to lift, they're doing all those things, but they didn't know what the actual problem was. And so the before and after is we helped quantify the data. We pointed them to that. I mentioned before those heat maps. They had a very specific problem on third shift where they were understaffed and people were trying to work fast to hit their productivity goals. And they had a it was people kind of breaking normal process and pushing containers underneath railings and other gentlemen putting their arms overhead and receiving it. And guess what? That really crushes your shoulders. So we pointed them to a specific area of need and problem. They changed how they thought about staffing, they changed how they thought about productivity goals, uh, but not decreasing those, uh, actually increasing their throughput, but at the same time massively decreasing their number of ergonomics claims or number of injuries or number of loss days. Um and it goes back to like they knew they had a problem, but they didn't know where and how to address it. Another example that I think is even faster that was interesting, um, another customer with a lot of uh pretty serious vehicle incident, uh, where their workers lost a leg and they were concerned about vehicle safety. Um, we turned it on, and within about a week, we we uncovered two things. Uh, one, someone had changed the governors of their forklifts, and so they were driving faster than what the management thought they were driving. Goes back to the point about noise. We had over 2,000 speed incidents on day one. First, we thought the algorithm was wrong. We're like, oh no, something's wrong with our product. And then we were like, no, no, they're governors. Someone changed the governors of their forklifts. Uh, so that was one thing. It was very quick change, very easy to address. Uh, but the second problem was they had a lot of employee turnover. And what was happening is they had a lot of new employees who were coming on the forklifts in high intensity environments and immediately just going to work in areas where there were narrow choke points, there were lacks lack of mirrors. And so we worked with them on a couple of things um, better training on. Practices for forklift safety, but at the same time, some environmental recommendations on more visible convex mirrors and more completely demarcated floor lanes. They saw a 93% reduction in reportables related to vehicle safety in the first year. And they haven't had a serious injury or fatality since they implemented Voxel. So, like those are the types of things where it's like very quick solution, but also very long-lasting outcomes.

SPEAKER_00 22:26

And Vernon, you know, I'll talk about ROI in a little bit, but um it seems to me compared to a lot of technologies I've witnessed in my life, this would be easier to quantify the ROI. Um I know there's challenges with it anytime you calculate ROI, but uh and the and the amount companies are paying because of these safety issues is just enormous now. The size of them has grown. I read a Wall Street Journal article that was talking about the number of uh lawsuits and everything hasn't increased, but the size of the uh awards have increased dramatically. Um but I'll get to that in a minute. But I'm just curious. Um so I have in my mind a way you can calculate ROI, but I'm wondering it seems to me there may be some um additional unexpected operational benefits from using um a solution like this.

Safer DCs And Better Inventory Flow

SPEAKER_02 23:29

There are. Um I'd love to kind of combine my answer and maybe talk a little bit about ROI now as well as some of the benefits because I think they're they're inherently related. So the great thing about the product is it is worker friendly. We're making them safer. It's leadership and culture friendly, something that that leaders and and they do care about. Sometimes they don't know how to fix it, but it's also finance and bottom line friendly. So very demonstrable ROI. Here were your claims pre-voxel, here are your claims post-voxel, here are your OSHA finds pre-voxel, here are your OSHA finds post-voxel. Um, and so here are your lost days pre and post. And it's pretty easy to measure because we've we're so laser focused on the types of problems we address. And so we can say your ergonomics claims went from X to Y, your vehicle incidents went from A to B. Uh, and really be able to quantify that. We typically tell typically tell customers they can expect a three to four X ROI measurable in the first year. Um, again, getting kind of changing the notion of software where it used to be like, oh, by year three, you're gonna save a ton. No, we're saving customers money right away. But two really cool things happened um in parallel. Uh so one of the unintended benefits, we've done a few case studies with our customers where they've actually seen a significant decrease in employee attrition uh for the sites where they had voxel, where they didn't. And it makes intuitive sense as soon as you unpack it, right? So, by way of my quick background, I grew up in Indiana, my family all works in very working class sorts of jobs, manufacturing, agriculture, first person in my family to have this sort of job and this sort of career. And so this is something I know really well. You don't want to work with people who are driving like maniacs in the warehouse. Like, and everybody knows who they are, right? Like, so you're you're able to like feel safer going to work. You're working with people who are acting and behaving more professionally. You have a company that's showing and investing in technology that is human-oriented versus automation and replacement and punitive-oriented, right? We actually bore faces. It's not about punitive, it's about positive uh reinforcement. And so this drop in attrition is really exciting, particularly when you think about the competition for skilled labor in the heartland of America as we try to reindustrialize. There's still a significant shortage of the types of laborers that these facilities need. And so if we're increasing, or sorry, increasing employee retention, what an awesome outcome, right? One that just makes really, really intuitive sense. Uh, the second one, and this is one of our big customers that said this to us, they're like a clean and safe environment is a more productive environment, period. Because safety issues are usually from unproductive uh issues or cleanliness issues or challenging, things like that. And so we've we've done a few case studies, none of them have been published yet. They're we're working through that. Where um a large cargo handling company, this one will be coming out in a couple of weeks, they've seen about a 15% increase in throughput. At the same time, they've seen a 60% decrease in recordable injuries. And and not don't get me wrong, they're focused on throughput, so they're doing other stuff. But technologically, they haven't really made other major AI investments. And so, like a piece of this productivity is the fact that by creating a safer environment, we've made sure that workflows are better, we've made sure that process compliance is higher, we've made sure that the messiness that leads to sort of mistakes and accidents is gone, which is increasing our labor's ability to get stuff done and doing it safer. Um, I've always detested and get really frustrated when people are like, oh yeah, you can make everything really, really safe, but but you know, ultimately we have to run a business and it's bottom line oriented. That's an antiquated, like 1920s way of thinking. You can be both safe and efficient by the proper use of technology. So those are some of the really cool outcomes that are happening on top of just the that that measurable bottom line uh ROI.

SPEAKER_00 27:22

I was even thinking, this is what one thing that came to my mind, and part of this is just because of all my work in inventory management and forecasting. But um, you know, if you have a DC um and you like in your case, um you'll are seeing 65% reduction in claims, well, there's probably even more um there's probably even more um safety issues that had been reduced than you see the claims. In other words, the you know, the there's this many safety issues and there's this many claims, but they're not the same. Um but anytime you have an accident, it it can throw things off. You know, you you might not send a truck on time, you might not send it uh you know full uh with everything that it needed. And so if you've got a DC that's seen that level of a reduction in accidents, then the the re the the customer of that DC could be retail store, could be whatever, uh I would think would need to hold less safety stock because you've reduced the uncertainty you know in the supply and in the lead time. Um I I know that would be hard to calculate, but I think it's a really important variable because if you have a stock out in the store, you if you have a stock out in the store because of a safety issue in the DC, your cost of a lost sale is much higher. You know, uh so on the one hand, you're gonna have fewer lost sales in the store. On the other hand, you're not gonna have to hold as much safety stock to deal with these kinds of uncertainties. Do do people think about that, customers at all? They do.

Mission Driven AI And Closing

SPEAKER_02 29:19

It's it's a really good point. And interestingly, uh, you know, we work with five of the top 20 retailers in the United States. Uh and that cohort, they have they know that. Like they they've already done all the work internally. They're like, we understand that these sorts of safety issues have an impact too, to your point, improperly filled trailers, uh, you know, not full trailers, all kinds of myriad issues, which then downstream impacts revenue. So they're telling us, like, when we go to them, we're like, oh, here's the ROI. They're like, no, no, no, we we don't we don't need you to do that. We already know that this is a massive problem and that if we can fix safety issues, it's like X mini hundreds of millions and more in terms of bottom line and top line. One of the things that we're we're working on is how do we help that next tier, right? The ones where they don't have their own internal BI departments, their own, you know, internal AI uh teams that have built out these models who they're like, we know the value. Um, how do we help them understand that value? Because there's there's so much to it. It goes back to the point about safety as an outcome of operations. The more operational improvement you make, the safer you are. And if you invert that, if we can help point to where inefficient operations lead to safety issues, it can obviously lead to really good outcomes from about a top line perspective. We just haven't quantified it yet, Matt. So we hear it, we know that it's there. Um, and that's certainly something we want to be be really focused on in terms of like the next generation of where we're going from both a product perspective and a storytelling perspective.

SPEAKER_00 30:54

Well, Vernon, I I've really enjoyed this conversation. And um, of course, watching Voxel now for for some time, um, I must say congratulations on the amazing job you've done leading Voxel um in so many ways. And and really this interview just kind of confirms uh why you're really good at articulating the uh the benefits and the challenges. And um so congratulations.

SPEAKER_02 31:25

Oh, thank thank you, Matt. I I um I really appreciate that. I I would just say maybe finish it on this. One, it's amazing to be building the type of AI that's both cutting edge from a technological perspective, but that has a human orientation. Um, and the cool thing about that is not only do you get to feel good about what you're building, I'm a very mission-oriented person. You can attract awesome talent, right? So we're going out and bringing in the best engineers from Carnegie Mellon and Caltech and all these places who then worked at, you know, Waymo and Aurora and all these really cool autonomous vehicles and companies because they're getting to do something that's a mission outcome with leading edge technology. And to your later point, it's not some abstract concept that's going to take five years. They can feel it right away. Uh, and so it's really that combination of being able to do something that that feels good and is good while also still kind of appealing to that you know innovator and and creator mindset. Um, it's a total team effort, and I really appreciate you having me on. And uh I love talking about voxel anywhere and anytime I can.

SPEAKER_00 32:32

Yeah, that was it's it's been enjoyable. And I can, to your point, I can see uh, I mean, you're saving people people's lives, you're um saving them from injuries that are painful and uh uncomfortable for years. Um, so very very impressive. Well, again, good talking to you, Verna, and thank you for taking time to join me. I really appreciate it. Thank you, Matt. I appreciate it.