From Olympic Trials Swimmer To AI Founder, Kaitlyn Albertoli's Mission To Protect Critical Infrastructure

[00:00:00] - [Speaker 0]
A big thank you to Denodo for helping me make more than 60 monthly interviews possible across the Tech Talks network. And as businesses move from GenAI to Agentic AI, trusted data becomes everything. Everything from GenAI to AgenTik AI, Denodo is helping organizations build intelligent, secure, and scalable AI solutions with data access, governance, and explainable results. So build AI that you can trust and do it with Denodo. And you can learn more by simply visiting denodo.com.

[00:00:40] - [Speaker 0]
What if the future of keeping the lights on in your home, in your organization depends on whether AI spots the problems we cannot see long before they become outages, failures, or even wildfires? Well, today, I've invited the cofounder and CEO of Buzz Solutions on the podcast to join me, which is a company bringing together drones, computer vision, and geospatial intelligence to help utilities inspect and protect critical grid infrastructure at an entirely different level. This is one of those stories that feels super timely for all the right reasons. Wherever you're living, you've probably seen increased risks around the weather, aging infrastructure, technical debt, and and all this comes at the time where we keep piling more pressure onto the grid, especially with ironically AI. So the pressure on utilities and field teams has never been greater, and there's also a shortfall of talent in the industry as well that we need to recognize.

[00:01:51] - [Speaker 0]
And we'll learn more about how crews are turning to AI systems that can process massive volumes of imagery in hours and turn that into practical insights for the people doing the work on the ground. We'll also talk about solar, grid resilience, and why this technology is at its best when it supports human expertise rather than just trying to replace it. Real important one, this one. So enough scene setting for me. Let me introduce you to my guest right now.

[00:02:22] - [Speaker 0]
So thank you for joining me on the podcast today. Can you tell everyone listening a little about who you are and what you do?

[00:02:29] - [Speaker 1]
Yes. Thank you so much for having me. I'm excited to be here. I'm Caitlin, cofounder and CEO here at Buzz Solutions. At Buzz Solutions, we are a visual AI platform turning massive amounts of utility inspection data into prioritized actionable insights to help reduce wildfires, power outages, and forced shutdowns from failed grid infrastructure.

[00:02:52] - [Speaker 0]
And not only that, it's incredibly cool, the work that you're doing here. You must have one of the coolest jobs in the world because you're working right at the intersection of AI, drones, critical infrastructure. But for everybody listening, wherever they are in the world, can you paint a picture of of just how power grid inspections have traditionally worked and and what changes when you introduce computer vision, geospatial intelligence, and and so many other cool technologies?

[00:03:18] - [Speaker 1]
Absolutely. We have certainly seen power utility inspections shift tremendously over time. So utilities are mandated to inspect a portion of their infrastructure every year. Typically, it's about 20% of their infrastructure per year, which means over a five year period, they will cover their entire grid system. Utilities are typically responsible for managing all of their physical grid infrastructure.

[00:03:45] - [Speaker 1]
That includes their towers that you see, their poles that you see in the cities, their substation infrastructure, and even generation infrastructure as well. Historically, utility linemen and field crews would actually walk the lines with binoculars and a pen and paper to look for any defects they could see from the ground. Of course, there are so many different defects that you can't see from the ground that require some sort of an aerial inspection. Then they move to helicopters. So with helicopters getting that top down that top down view, then they would still have the person with the binoculars and a pen and paper writing down any sort of critical defects they would see.

[00:04:26] - [Speaker 1]
But recently, in the last five to ten years, we've seen the introduction of drones and fixed wing aircrafts really shift the landscape. It took utilities from collecting maybe a couple images from helicopters to hundreds of thousands and millions of images per year. Well, if you think about that volume of data, you can no longer sit there and manually analyze with the pen and paper. And so that's ultimately where, we've come in with with our solution at at Buzz is to help automate that process of the very manual data analysis piece. And one more thing I'll mention that's quite interesting is because utilities are now starting to adopt more of these autonomous technologies, like autonomous drones and autonomous fixed wing aircrafts, they're actually able to do more than 20% of their assets inspected per year, which means they can move from a very reactive maintenance state or, you know, only getting eyes on a certain portion of their territory to a much more proactive state where they can understand how their infrastructure is performing over time, not just every five years, but maybe they can inspect it once every two to three years, or the goal eventually would would even be once a year.

[00:05:37] - [Speaker 0]
I love that. I also wanna give a quick shout out to deskless workers because they comprise 80% of the global workforce. That's 2,700,000,000 people that are doing things like site inspections. And for the large part, technology has not been focused on that. The technology focused on the other 20% of people sat at desk.

[00:05:57] - [Speaker 0]
And I think it's such an important thing here because these are the people that are protecting our critical infrastructure and ensuring that the maintenance, etcetera, is carried out. And, Bus Solutions, you can process tens of thousands of inspection images in hours rather than weeks. So what does that speed actually unlock for utilities and teams on those deskless workers in terms of decision making, cost, and risk reduction? Because it there's a real ROI here, isn't it?

[00:06:27] - [Speaker 1]
Sure. And first off, I wanted to just, underscore the point that you made about the deskless workers and the importance of those frontline workers. I mean, we see today that there is such a, labor shortage of linemen and field technicians who are critical, workers in order to keep the lights on, in order to keep our power infrastructure maintained. And so, the frontline workers are just so critical and so important to the success of, you know, keeping our our power up and our grid online. But yes.

[00:06:59] - [Speaker 1]
So at at Buzz, we take, all these different visual, visual data inputs, largely including, images and even video feeds, historically, which would have taken months or maybe even years to manually analyze them. We're able to take, you know, hundreds of thousands of images and analyze them in even less than ten hours. And we do that through our proprietary machine learning computer vision AI algorithms that we've built and trained, over the the large number of of years with proprietary utility data. And the reason that we're able to, analyze these images so quickly is we have pre trained models that are really tailored specific to utility infrastructure. And so we're able to work, to do asset inventorying detections for all the different components on a tower or on a pole, and then we're able to also look at the condition of those assets.

[00:07:55] - [Speaker 1]
So we're then triaging the most critical defects that happen, maybe like a rusted c hook, which could cause a downed line or a, splitting of damage of a cross arm that could cause a a cross arm that could compromise the cross arm's integrity. Or maybe it's, looking at insulators, which are flashed or broken or cracked. And so we're able to take these large amounts of data, sift through them with our algorithms, and just pick out the most critical defects that utilities need to address right away. The remainder of those defects, we then triage based off their severity. And at the end of the day, we actually hand the decision making power to those frontline workers to say, yes, this needs maintenance.

[00:08:39] - [Speaker 1]
No. We can wait six months and track this issue so that they can then optimize where they're sending crews to the field, again, to help them reduce their maintenance backlog and be able to be more effective in where they're sending those field crews.

[00:08:53] - [Speaker 0]
Before we started recording today, we're both talking about the weather and how how much it's changed, and there's, what, 5,000 miles between us talking today, and we are all seeing weather related threats becoming a year round concern. So how is AI helping utilities move from reactive maintenance to actually predicting failures before they happen, especially when it comes to anything from wildfires to vegetation risks?

[00:09:19] - [Speaker 1]
It's a great question. And maybe to to look at it in a couple of different ways. So let's take wildfires first. You know, with wildfire, risk mitigation or wildfire planning, you really need a a two pronged approach. You need early detection of a wildfire that's already started and more preventative measures for, prior to a wildfire starting.

[00:09:40] - [Speaker 1]
We're on the preventative piece of it. And so, again, as we were talking earlier, historically, you you've seen utilities inspecting 20% of their infrastructure per year. However, as we've seen grid infrastructure is continuing to age, many components that are deployed out in the fields and many assets deployed out in the field are actually beyond their end of life, which means today we're working off borrowed time of, components that are are past their shelf life or are past their expected expiration date. And what we're seeing is with the introduction of, you know, vehicles coming online, renewables coming online so quickly, and then as we've seen in recent years, data centers onboarding. All of that additional, all of that additional load growth is causing even more stress and pressure and strain on those physical components.

[00:10:32] - [Speaker 1]
The problem with that is you have this kind of perfect storm of factors. You have aging grid infrastructure, components that are past the end of life, then you have additional stress and strain. And then you add on top of that all the weather related stress and strain that's being posed as well. It puts the utility assets at a real, risk, you know, at a much greater, much greater point of risk today. So, in order to identify where you need to send maintenance crews or in order to identify where you need to conduct that maintenance, you have to get more eyes more consistently on those components.

[00:11:09] - [Speaker 1]
And with a solution like like ours, we're able to track those assets over time. So we're analyzing, all of this large volume of imagery, not only just to pick out the specific pain points or the critical pain points, but we're also tracking the assets which may have been good. There may have been no issue with them in the past. And then we're able to track, okay. We see early signs of some sort of damage happening here, and then we're able to track it over time, leading a utility to be able to more proactively, replace that that equipment or replace that, that asset or that component prior to it failing.

[00:11:45] - [Speaker 1]
So that's one thing that's particularly important on the wildfire side of things is being able to identify, one, the most critical areas of risk that you have on your grid, areas where there's stress and strain. Two, being able to get eyes on it ahead of it failing. And three, being able to send those maintenance crews more effectively, to that zone, especially ahead of wildfire season. Now I wanna shift maybe to talk about storms because that one's slightly different. And so storms, we see things like vegetation encroachment is a really big challenge with storms, trees falling into lines, vegetation coming in contact with the poles themselves.

[00:12:23] - [Speaker 1]
Prior to a storm, being able to get out and inspect that infrastructure to understand where you have key areas of risk, whether it's vegetation or whether it's a cross arm, which may not have adequate structure structural integrity, which may be more damaged. Being able to proactively get out there and upgrade that equipment or trim that tree helps the utility reduce their overall risk exposure. What we're also helping do is, by doing that baseline inspection, then the utility can clearly understand what damage was a result of the storm itself versus what may have been a preexisting defect or preexisting anomaly that existed on that pole. And in a post storm environment, every minute counts for getting the power restored. But in a post storm environment, you also have to inspect the infrastructure, prior to reenergizing the line.

[00:13:17] - [Speaker 1]
And so if you're manually, going out there flying this infrastructure, you're manually analyzing those datasets, of course, that's so much more time than if you're to use an automated solution like what we're able to provide here at Buzz. And so I I'd say in the post storm environment, speed is of the essence. And being able to have that pre storm inspection and that post storm inspection allows utilities to use that data much more effectively to, more adequately send their field crews to the right poles and towers.

[00:13:48] - [Speaker 0]
And when you're analyzing huge volumes of imagery, I'm curious. How do you ensure the system is spotting the issues that humans might miss without overwhelming those human teams with false positives or adding to that alert fatigue that they might be experiencing?

[00:14:04] - [Speaker 1]
That is a great question, and that's always a very fine line between false positives and false negatives. To take an example in substation security, so we actually do a lot of of substation security work as well, looking for intruders, whether that's a person, whether that's, you know, a a car, whether that's an animal in a substation, we're able to alert to all those types of detections. But you can imagine if you're over alerting and you're sending maybe every single instance, let's say, a plastic bag that's floating by or some sort of a a tumbleweed or something like that, if you're over alerting, very quickly, the field crews will stop paying attention as closely to those datasets. And it almost becomes a boy who cried wolf type, environments if you're over alerting. And so that's a a real example if you just wanna be alerting to the the critical anomalies.

[00:14:56] - [Speaker 1]
If you have someone entering your substation, you certainly don't wanna miss that as well because that's a critical security risk in your substation. And so it requires a a fine balance of the identifying false positives and false negatives. And so, we spend the first, probably month to six weeks during what we call our implementation timeline, identifying what's what matters most to the utility. What's their tolerance for false positives? What's their tolerance for false negatives?

[00:15:24] - [Speaker 1]
And then we'll continue to tune more more specifically, and provide a more tailored, I guess, say, detection detection framework for that specific utility. And that becomes a really honest conversation we have to have upfront because the AI is never going to be a 100% accurate, but we certainly want to reduce the time that someone has to manually interface with the tool. And so what typically happens is we're looking at the, lower, false negative rates so that we don't miss an anomaly for the most critical defect detections, And there may be a little bit more willingness to have a higher false positive rate just so that we don't miss a critical defect. Whereas on the other side, we may be optimizing for the opposite. And so it requires an honest conversation with the utility, but it's certainly something that differs by different detection types and by kind of the importance of each of those detections that we're providing.

[00:16:22] - [Speaker 0]
And you're also partnering with companies like Skydio and Esri. So how important are these to your ecosystem approach when building AI solutions for something as complex and equally as regulated as a a power grid, for example? Is that so much going on here? But tell me more about that those partnerships and that ecosystem.

[00:16:43] - [Speaker 1]
Sure. Well, maybe just start out. It's worth taking a look at one of the big challenges that the utility infrastructure or the utility industry, excuse me, is facing today, and that's data siloing. Data siloing is a huge problem that the industry is facing more broadly. To provide an example, you know, several of our utility partners that we're working with are collecting data from maybe five, ten different, sensor types or different, inspection vehicles.

[00:17:13] - [Speaker 1]
That could be drones. That could be helicopters, fixed wing, ground based, static cameras. It's really a blend. Historically, all of those datasets have been stored separately. And so the utility hasn't been able to leverage multiple of those data sources, together to be able to drive the most comprehensive insights and best, best output from the inspection that they already invested in going out to send crews to go conduct.

[00:17:43] - [Speaker 1]
And so, a solution like ours starts by helping utilities break down that data siloing. And to get to your your point of ecosystem partners, partners, that's exactly where ecosystem partners are so helpful for us. So to talk about the GIS piece of it, we work with so many utilities who have set up, these great Esri environments, but it's a really manual process for their teams to go through and update that system of record. So let's say on the distribution side, they may have had a big storm come through and they had to replace a thousand or a couple of thousand assets after that storm. Oftentimes, that data of those new poll locations won't make its way back into GIS.

[00:18:26] - [Speaker 1]
So when the utility field crews go out to inspect that infrastructure again, they may have a misalignment in their GIS historical data and where the polls actually are today because of that data siloing, the system of record wasn't updated. And so oftentimes, that's where we start with utilities is saying, okay. We recognize that the GIS data is likely not perfect, but we wanna get you to a place where you're continually updating that GIS data so it becomes your single source of truth. That can help you then send more autonomous drone inspections because you have all the right poll locations, and it also can help save your field crews a tremendous amount of time in manual data entry. And so we spend a lot of time automating that process to, save the manual data input timing.

[00:19:16] - [Speaker 1]
And then the same thing exists with, Skydio. We have an integration directly with the Skydio cloud, so we can pull data directly from, the utility's inspections from the Skydio cloud. We can pull it into our platform, analyze it, process it, and then push it to the relevant platforms that the utility, would like to send the data to, whether it's a cloud storage bucket, whether it's, again, GIS, asset management, work management, and that helps streamline that, again, manual process historically. The utility no longer has to, you know, download that data, upload it to our platform. We can bring that connection together.

[00:19:53] - [Speaker 1]
And so helping reduce data siloing across the board has been another significant value for many of our utility partners too.

[00:20:01] - [Speaker 0]
And before you join me on the podcast today, I was also reading how you've started expanding into solar inspections too where small efficiencies can quietly add up. So what are you learning about how AI can improve performance, reliability, etcetera, in in renewable energy systems? Any big takeaways there or any big lessons learned?

[00:20:23] - [Speaker 1]
Absolutely. Solar is a very interesting space in part because we're seeing so much build out of new solar on the utility scale solar side of things. Many of our utilities that we're working with are five to seven to 10 x ing their solar their solar builds out over the next five to seven years. However, they have the same field crews inspecting and maintaining the solar assets even as they're scaling their solar their solar fields by that much, which puts them in a a big, I guess, should say, bottleneck as they are trying to inspect more of that infrastructure and, be able to conduct more prioritized maintenance. And so our solution helps them inspect more regularly, not only in an ongoing maintenance environment, but also just after the new panels are conducted.

[00:21:13] - [Speaker 1]
We often see that there's damage that happens during the construction process and as it's handed over to operations. And so we're able to identify what issues existed post construction so the utility can quickly correct those issues. And then they have a true baseline health of the solar panels on day one that then they can benchmark against over time. One common use case that we've seen is when a panel will fail, prior to warranty expiration, oftentimes if the utility doesn't get out there to inspect it with great regularity, they may not identify that that solar panel may have failed prior to warranty expiration. That's a huge lost, a lost cost for them and an opportunity to repair that panel prior to the warranty expiring.

[00:21:59] - [Speaker 1]
And so that's just one example of how more frequent inspections and insights like ours can help them make more proactive, not only maintenance, but better business decisions as well.

[00:22:10] - [Speaker 0]
I love that. And on a personal note, I noticed that, when I was doing a little research on you, you've got quite the backstory, a great origin story as well. You're someone that competed in the twenty twelve US Olympic trials in swimming. You've gone from there to leading a fast growing AI company. I I'm curious.

[00:22:28] - [Speaker 0]
How has that mindset shaped the way you build a team, approach challenges, and scale a business in such a high stakes industry? Are there any synergies there when you look back at your career?

[00:22:38] - [Speaker 1]
Thank you. Yeah. It's it's actually very interesting. There are many similarities between, you know, highly competitive athletics and and swimming specifically and, running running a startup. I remember people used to tell me when I was swimming all the time, they said, you're going to appreciate the time management one day that the skills that you developed from swimming.

[00:23:00] - [Speaker 1]
Although, don't think that was the most important, skill set that I was able to develop from from swim. I think actually it was the resilience because, I was a sprint freestyler. And so in sprint freestyle, everything comes down to hundreds and really even thousands of a second if you think about it that can separate you from, you know, 20 spots lower on, on the rankings. And so it comes down to just these very minor tweaks that you're having to make. But, ultimately, I'd say resilience is really critical because after one race, you have to get back up and go out and swim another.

[00:23:36] - [Speaker 1]
Whether it was a a good race, a bad race, you know, you're often swimming multiple races a day, you know, across almost every weekend. And the resilience that it takes to be able to course correct very quickly, put yourself right back in the the right mindset to go out and and do it again and be able to get back up and keep going is is really important. And I'd say that's the same in entrepreneurship. I mean, and it's the same in an AI company is we've seen the AI landscape and the utility landscape has changed so much over the last five years. I mean, even since the time we've been in the market, there have been so many shifts and changes in how utilities are thinking about AI, how utilities are thinking about inspections, even how technology has changed in how we build, construct, and deploy AI models.

[00:24:24] - [Speaker 1]
And so the ability to understand, you know, when a challenge hits, when, you know, things aren't going your way, how you can get back up, course correct, and keep pushing through is just really important, I think, in the entrepreneurial journey.

[00:24:37] - [Speaker 0]
Wow. What an incredible story, and I love what you're doing here. And for anyone listening wanting to a little bit deeper on anything we talked about, where would you like me to point everyone listening?

[00:24:48] - [Speaker 1]
Absolutely. So, we're very active on LinkedIn. You can you can find me on LinkedIn. It's Caitlin Albertoli. You can also find Buzz Solutions on LinkedIn.

[00:24:57] - [Speaker 1]
It's just Buzz Solutions, and same thing on our website. We have a lot of new and exciting announcements coming out about some of our utility partners that we're working with, some of our case studies. So definitely be sure to check those out. And then we talked about solar today as well. We are just announcing our latest solar products and solar offering in the space.

[00:25:16] - [Speaker 1]
So definitely, check check that out and learn more either on our website or on our LinkedIn.

[00:25:22] - [Speaker 0]
Well, looking at all the technology you're work working with here and AI powered visual intelligence for utilities, partnering with drone providers. All sounds incredibly cool. But what I love more than anything is at the heart of it all, you're talking about enhancing human expertise, those deskless workers not replacing it. So I will have links to everything that you've mentioned there. I'll be following your story very closely.

[00:25:46] - [Speaker 0]
It'd be great to get you back on next year seeing how things are evolving. But, Monethy, thank you for joining me today and sharing your great story. Thank you.

[00:25:54] - [Speaker 1]
Thank you so much for having me. I I really appreciated the conversation.

[00:25:58] - [Speaker 0]
So what really stayed with me from this conversation is that this is not AI for the sake of a headline. This is AI being used where it genuinely matters, helping utilities finding problems faster, prioritizing risks better, and supporting the field crews responsible for keeping all of our critical infrastructure running, that infrastructure that we're probably guilty of taking for granted. And my guest brought a very human perspective to this conversation because behind all the drone, AI models, and analytics was a very clear message about resilience. Whether that means building a company, responding to a storm, or helping utilities shift from reacting to failures to preventing them in the first place. And all this feels like my big takeaway here, because the smartest use of AI is often the one that helps people make better calls faster, and do so in moments where timing really matters.

[00:26:58] - [Speaker 0]
And if this is where grid inspections are heading, certainly feels like an important and welcome shift. So I'll include links to everything so you can find more information about my guest and the work that she's doing. And please pop by techtalksnetwork.com if you wanna browse through 4,000 interviews, work with me, see me in person at an event near you. Just let me know, and I will speak with you or see you then. A quick thank you to NordLayer for supporting the podcast and helping me make these daily conversations possible.

[00:27:30] - [Speaker 0]
And if you are listening and you're responsible for security or IT, you will know the reality. The reality that most of your risk now sits inside SaaS apps and browser activity. That gap is exactly what NordLayer is addressing with its new business browser. So instead of bolting security on from the outside, it builds it directly into the browser itself. This means you can control access, monitor activity, enforce policies, and reduce shadow IT all from one single place.

[00:28:05] - [Speaker 0]
And most importantly, it does it without adding deployment headaches or complex onboarding. You get things like browser based data loss prevention, SaaS access control, and zero trust browsing, but delivered in a way that your team can actually use. So if you've been trying to simplify your stack while improving visibility, please check it out at nordlayer.com/browser. But that's it for today, so I will speak with you all again very soon. Bye for now.