AI transcript
0:00:15 Hello, and welcome to the NVIDIA AI podcast. I’m your host, Noah Kravitz.
0:00:20 Buzz Solutions is on a mission to enhance the safety and efficiency of the electric grid
0:00:25 through innovative data analytics and machine learning. A member of NVIDIA’s inception program
0:00:29 for startups, Buzz’s visual intelligence empowers utility companies to better monitor
0:00:34 and manage their infrastructure. The result is improved reliability and reduced outages.
0:00:39 Here to explain how Buzz does it and to talk about the impact AI can have on making our electric
0:00:46 grids safer and more robust is Caitlin Albertoli, CEO and co-founder of Buzz Solutions. Caitlin,
0:00:51 welcome, and thanks so much for joining the NVIDIA AI podcast. Thanks so much for having me, Noah. I’m
0:00:56 excited to be here. So we’re two Californians, and I mean, I think it’s everywhere now, but certainly in
0:01:01 our state, the electric grid has been a source of news for quite some time now. It’s something
0:01:06 that’s on lots of people’s minds. So I’m very excited to learn more about what Buzz is doing
0:01:12 and the impact that AI can have going forward to really help all of us because we all use electricity,
0:01:18 right? So maybe we can get into it with a little bit about your background and how you came to co-found
0:01:24 Buzz Solutions. Absolutely. Thanks for the question. Happy to share. So we launched Buzz Solutions from a
0:01:32 launchpad course at Stanford University in the spring of 2017. My background is actually not from the electric and
0:01:38 power industry. Prior to Buzz, I was working in finance, and I also ran a nonprofit in the sustainable food
0:01:45 world. Before that, I grew up in Southern California in a town that was close to the nuclear power plant, the
0:01:51 San Onofre power plant. So that was really my first exposure to the power world, the power industry. It was
0:01:57 actually shut down during the time that I was living in Southern California. So that was an interesting dynamic
0:02:02 to, you know, grow up with and see. Yeah. Did that leave an impression on you? I mean, did it, did you think more
0:02:07 about where power comes from than perhaps, you know, the average kid? Definitely. It definitely got me thinking
0:02:14 more about our different generation sources and renewable energy more broadly. And so when I went into Stanford, I was
0:02:19 certainly looking for more opportunities to learn about about generation and power as a whole.
0:02:24 That’s great. So there’s a lot of steps on the way, as you mentioned, with, you know, finance and sustainable
0:02:28 food. But how did you wind up co-founding Buzz? How did Buzz come to be?
0:02:35 Sure. So my co-founder, Vic, and I met in this launchpad course that was in the School of Civil and Environmental
0:02:40 Engineering. It was about building a startup in the School of Civil and Environmental Engineering focused
0:02:46 on energy and sustainability. And there was something that called me to this class because I was always
0:02:52 really entrepreneurial and interested in learning more about starting a business. But the, I guess,
0:02:57 the intersection of energy and sustainability, given my background, with entrepreneurship seems like
0:03:03 a really unique opportunity. And so that was the course that, that we ultimately met. And when we
0:03:09 were originally exploring ideas for, you know, what we wanted to, I should say, not really found a company
0:03:15 at that time, but what we wanted to use for this class, we were looking at the wind turbine market. We were
0:03:21 looking at all that was happening with optimization of placement of wind turbines and how you could use
0:03:26 drones and technology to understand where it was best to place those wind turbines. But in those
0:03:33 conversations with some of those, you know, those wind turbine companies and manufacturers, we very
0:03:37 quickly were tipped off to the power space. You know, everyone said, this is a huge market and
0:03:43 there’s definitely an opportunity in wind. But have you seen what’s happening with inspections in the
0:03:50 power space? And kind of an interesting time, because if you think about 2017, that was before a lot of the
0:03:55 big wildfires, before a lot of the power outages, and before some of the major blackouts that, of course, we know
0:04:00 about today and what has unfortunately transpired since that time. But we ended up taking that
0:04:06 conversation and leverage the Stanford alumni network to interview 35 investor on utilities. We were able
0:04:13 to get conversations with 35 investor on utilities. Can you imagine that that many? And they all had the
0:04:19 same story. They were collecting a lot more data, or they had plans to collect a lot more inspection data,
0:04:24 whether it was due to, you know, improve technology that allowed them to do more inspections, and also
0:04:29 regulations that were driving more inspections. But they didn’t have a way to manage or analyze that
0:04:33 data. And so that was really kind of the genesis of ultimately how we launched Buzz.
0:04:38 Yeah. So Caitlin, it’s funny, because I’m thinking about that kind of pre-pandemic time and going to
0:04:46 NVIDIA GTC back, you know, probably around 2018, 2019, maybe 2017, and watching a demo of drones and
0:04:52 kind of a, you know, is it a demo in the convention space, but a drone in a simulated warehouse environment
0:04:57 flying around doing inspections and doing what’s it called pick and pack and that kind of thing.
0:05:02 So what you’re talking about rings a bell for me, if in a different space. But how did you, you know,
0:05:06 when you’re talking about there’s all this data, what do we do with it? And to me, that says AI,
0:05:11 but this is, you know, seven, eight years later. So back then, did you see instantly, did you see
0:05:15 all the data and think, oh, machine learning, AI, we can do something with this? Or kind of how did
0:05:20 that leap to AI happen? It’s a really good question, because at that time, for a lot of utilities,
0:05:26 it was still just an idea, or it was still just a plan that they were looking to put in place. You know,
0:05:31 they were buying their first drones, they were training their first pilots. And on top of that,
0:05:36 they were still doing, and present day still are doing, a ton with helicopters. So they were
0:05:41 starting to actually collect images with helicopters at that time, while also building drone programs.
0:05:47 And so even in our first days of Buzz, a lot of our conversations with utilities revolved around,
0:05:52 okay, we’re starting to use drones, we’re starting to collect data with drones. How do we scale this
0:05:58 program? How do we collect data in the best way that’s going to make AI the most successful,
0:06:03 I guess you would say? And we still have a lot of those conversations today. But data wasn’t
0:06:08 standardized in that space, and it still isn’t standardized. But if you can think in 2017,
0:06:14 that a more traditional industry like utilities, where they had five to seven year plans to scale
0:06:19 these massive drone programs, I mean, it’s lightning speed for some of these more traditional industries
0:06:25 like utilities. So it was more an idea at that time, but they had all of the right, I guess you would say,
0:06:29 pieces to the puzzle to be able to build some of these programs.
0:06:34 And so what’s happened along the way, and I want to get into, you know, some of the use cases,
0:06:39 some of the challenges and problems that you’ve solved and are helping energy companies solve,
0:06:44 but I don’t want to skip over the growth in between. Walk us through that. And if there are some of those,
0:06:49 you know, use cases and things that kind of popped up and are still prevalent, definitely dig into those if you would.
0:06:56 Sure. It was an interesting time back in the 2017-2018 timeframe. We’re talking about these concepts like
0:07:01 artificial intelligence, AI, and a lot of our user base at that time, they weren’t really actively using
0:07:06 software platforms. A lot of them were still manually drafting inspection reports. So when you’re talking
0:07:10 about going from manually drafting inspection reports to using AI…
0:07:12 And these are the utility companies you’re talking about.
0:07:14 Correct. Yeah. These are the utility companies.
0:07:16 Yeah. Digital transformation. Yeah.
0:07:23 Exactly. Lineman, field technicians, engineers, a lot of them were still manually drafting inspection
0:07:28 reports. And some of our, you know, lineman and field technicians that were using our solution,
0:07:32 we’re having to go through, you know, thousands of images, tens of thousands and hundreds of thousands
0:07:38 of images manually before they’re looking to use a tool like this. And so the reason I give you that
0:07:45 background is in order to show up to a utility and ultimately, you know, bring value with a solution
0:07:51 like ours, we had to plug into the utility workflow. And so from day one, when we were working with
0:07:56 utilities to build this solution, to train our algorithms, to build our platform, we were really
0:08:02 utility centric. We wanted to make sure that the algorithms that we were training were very closely
0:08:07 aligned with how a utility subject matter expert would have labeled this data manually.
0:08:14 Is that, or was that back then sort of standardized or was it that, you know, from company to company,
0:08:20 even within a company from sort of inspector to inspector, the format of those reports was just
0:08:22 sort of a free form, I guess.
0:08:27 I would say there’s, there’s definitely a lot of subjectivity as it pertains to how, how lineman or
0:08:33 field technicians would label the data and how they would classify a significance, as we call them,
0:08:39 like anomaly or should say a physical defect on a component is how someone could think of it more
0:08:44 commonly. But it’s, if you can see it, the defect on a component or the anomaly, that could be a little
0:08:50 subjective of how someone classifies it. But we see that there’s a pretty standard set of labels or total
0:08:56 classifications that a utility is looking for. And we worked with research organizations like EPRI,
0:09:01 which is the Electric Power Research Institute, one of the leading electric research institutes,
0:09:05 along with utilities to make sure that we got that nomenclature correct.
0:09:12 And when you’re talking about defects, anomalies, what kinds of things are, you know, are you looking
0:09:16 for? Are they looking for? I, I not knowing much about it all. I’m just imagining, you know,
0:09:21 there’s a squirrel chewing through a line or there’s a giant explosion. What kinds of things are being,
0:09:24 you know, captured as anomalies? Actually, squirrels do have a big impact.
0:09:30 So those are the ones I always hear stories about, you know, like, oh, what was the outage last? Oh,
0:09:34 is this squirrel at the substation again? Or, but is that really like they’re that one? Yeah, it is.
0:09:42 Actually in substations in particular, a substation is where a utility changes the voltage. So it goes from,
0:09:48 from highly energized lines to lower energized lines or vice versa. So the substation is the facility where
0:09:53 that that change takes place. A substation, a lot of our questions that we get from utilities are,
0:10:01 hey, can you do things like birds, squirrels, snakes? Snakes is a huge one. Coyotes, mountain lions. So
0:10:07 we actually get a lot of animal requests as well, because animals, if you think about it, you know,
0:10:11 they’re sitting on the infrastructure or they’re chewing through something. It could have a significant
0:10:17 impact on the equipment itself. And a person may not be out there to be able to see that,
0:10:22 but they may start to get a sensor that’s tripping. And they may say, oh, there’s,
0:10:27 there’s something wrong here, but they may not know what caused it. And so having the ability to
0:10:31 leverage visual data to say, oh, okay, there was a squirrel that came in contact with this or a snake
0:10:35 can be really helpful for that utility to make a better decision on maintenance then.
0:10:40 Just to get a sense of the scale that we’re talking about that you’re dealing with, the Buzz is dealing
0:10:47 with, what’s the best way to, like, how much data, how many square miles, how many, you know, data points
0:10:52 or images? How do you, what’s the best way to express, you know, the scale of what these companies
0:10:54 are doing and what Buzz is doing to help them?
0:10:59 Sure. Some of our utility customers are collecting hundreds of thousands and millions of images a
0:11:05 year. And that’s just as a part of their standard inspection processes. Utilities have various
0:11:10 different types of inspections. So they could be doing what’s called a flyby inspection, which is
0:11:15 just looking for anything major glaring that’s happening along the infrastructure. Or they could
0:11:20 be getting down to a comprehensive inspection where they have a drone that’s flying and taking multiple
0:11:27 pictures to look for things like a insulator disc that’s broken or, or a crack in a cross arm or a
0:11:32 woodpecker hole in a cross arm, or even something as small as a cotter pin that could be missing or
0:11:37 loose, you know, a bolt or something like that, that’s holding the insulators to the line. So there’s all
0:11:42 these different types of inspections that take place throughout the year. And a large utility may be
0:11:47 collecting millions of images of their infrastructure that they’re having to sift through on an annual
0:11:55 basis. We see that utilities are doing data collection with helicopters, drones, big swing aircrafts, ground
0:11:59 based data collection, some are using satellites. So there’s a variety of different ways that they’re
0:12:03 collecting this data. And typically, it’ll serve a specific purpose for that utility.
0:12:09 Got it. Okay. And then, so how does, how does buzz work? You have a couple of different products listed on
0:12:14 your website, I would imagine. Well, I don’t know if that’s the full scope of it or not. But maybe walk us through
0:12:19 some of the solutions that, you know, your customers, the utility companies are employing and kind of how
0:12:24 it works from, you know, imagining data collection and analysis and action, but that may be entirely
0:12:30 wrong. So how does it work? Sure. Great question. So at Buzz, we are here to modernize energy infrastructure
0:12:37 with actionable intelligence. The central focus of what we do here is to help utilities take mass amounts
0:12:43 of raw data and use AI, machine learning, computer vision to turn it into actionable information or
0:12:48 non-actionable information if it’s not significant enough for utility to actually take action on it
0:12:54 right away. And what that means for us is we’re taking in all these different visual data sources.
0:12:59 So as you’ll see on our website, we have two different products that are listed. Our first one is more
0:13:04 inspection oriented. It’s called Power AI. And so that’s where we’re taking in all of this inspection
0:13:10 data that I’ve been talking about from the drones or helicopters, ground-based. It’s really linear
0:13:15 infrastructure data that we help the utility by managing it, meaning we’ll ingest all that data.
0:13:21 They can upload it to our platform. We’ll help them map it so that they know exactly where all the images
0:13:26 are on their corridor. And then we’ll analyze it with our machine learning, computer vision algorithms.
0:13:34 What was really important for us as we were training our algorithms and before we ultimately deployed was
0:13:39 that we could show up to a utility with pre-trained algorithms, meaning on day one, we could start
0:13:45 analyzing their data and start delivering immediate value. So our Power AI platform has pre-trained
0:13:50 algorithms. We’ll analyze all that data. And for us, we can analyze an image in a fraction of a second.
0:13:56 So it’s really quick, really fast. And then once those images are analyzed, we display them on a
0:14:03 dashboard for a utility to be able to go through and they can actually edit, adjust labels, add,
0:14:08 delete labels as they see fit. And then all of that gets pushed into an inspection report. We can deploy
0:14:14 that in their workflow a variety of ways, whether it’s directly into a work management system or to a
0:14:20 geospatial information system, which is a GIS system for mapping and managing their inventory.
0:14:25 We can deploy it in a variety of different ways for their workflow. So that’s Power AI. And then our
0:14:32 second product is called Power Guard. And that’s the one I mentioned kind of as we do animal detections
0:14:38 and all that type of thing. But Power Guard is a real-time or a near real-time alerting system
0:14:44 that deploys on a fixed camera. And so for Power Guard, we’re doing things, same component monitoring
0:14:50 that we’re doing with Power AI. But here we’re doing it as an alerting system in a substation or
0:14:56 in a facility where a utility needs more of those kind of eyes and continuous monitoring. So we call
0:15:02 Power Guard our continuous monitoring solution for security at a substation, whether that’s an intruder,
0:15:10 person, car, animal, squirrel, exactly, whether it’s safety. So if someone’s injured on the job
0:15:14 or they’re injured in their work, and then we do all sorts of component monitoring, whether that’s
0:15:19 overheating of components or physical degradation on those components that can be seen. And we send
0:15:24 those alerts 24-7 to the relevant teams at the utility. So that’s kind of our two different products.
0:15:26 It’s all vision-based as well.
0:15:32 Right, of course. So in a, I don’t know, like an emergency situation, a wildfire, a storm,
0:15:38 or something else that’s causing, you know, whether it’s a sudden unexpected blackout, or if a utility
0:15:44 company decides they need to roll out, you know, brownout blackouts, I can understand how Power Guard
0:15:49 obviously would alert, you know, there’s something’s gone wrong at this substation. But how do Buzz’s solutions
0:15:54 play into helping utility companies deal with, you know, these kind of real-time
0:15:58 emergency situations that are coming up?
0:16:03 So for things like Power Guard, we also do smoke detections. So we can detect smoke or overheating
0:16:04 or sparking issues.
0:16:05 Is that all visual?
0:16:10 Yeah. So we’re able to alert to any of those incidences that are happening, whether it’s within
0:16:16 the substation or at the utility facility, or even in the surrounding area. So we can alert to
0:16:21 any potential fire risk that’s happening there. On the Power AI side for inspections,
0:16:27 we’re identifying critical risk issues that could cause a wildfire before they start.
0:16:32 It’s preventative maintenance, basically. So we’re identifying, you know, for example,
0:16:36 rusted components that could ultimately rust out and cause a line to drop down.
0:16:43 We are able to do those types of detections on a transmission tower distribution pole. So a high
0:16:49 voltage tower or a low voltage city line, we’re able to do those detections. And then we also can
0:16:54 do sparking detections as well. So if there’s any sort of a physical component that could cause a
0:16:59 sparking issue, we’re able to detect that on the physical components. And then we do things like
0:17:04 vegetation encroachment detections. So we can detect where vegetation is encroaching on a line.
0:17:10 And where it’s particularly valuable for utility is we can say, hey, there’s vegetation encroaching,
0:17:16 let’s say on this distribution or a lower voltage city line. There’s vegetation encroaching on this
0:17:22 distribution pole. And you also have five different anomalies or defects that are on this particular
0:17:27 tower, this particular pole. This is a high risk pole that you should definitely be maintaining. So we do
0:17:33 those types of things to help with wildfire risk mitigation. And then on the storm side of things,
0:17:40 utilities are starting to deploy more vision-based inspections like drone technologies post-storm
0:17:47 to go out and fly corridors so they can understand what sorts of anomalies exist on the infrastructure
0:17:53 before they re-energize that line. And by using Buzz, they’re able to analyze and process through that
0:17:57 data really quickly. So they understand exactly what they need to replace prior to re-energizing.
0:18:03 So we’re able to save mass amounts of time and help direct the field crews where to go to upgrade
0:18:05 and restore that infrastructure post-storm.
0:18:12 That’s fantastic. I’m speaking with Caitlin Albertoli. Caitlin is CEO and co-founder of Buzz Solutions.
0:18:17 And we’ve been talking about Caitlin’s journey to starting Buzz, but Buzz’s journey in their work
0:18:23 using machine learning, using computer vision, other AI technologies to help electric grids,
0:18:29 help utility companies deliver power safely, more effectively, do preventative maintenance,
0:18:33 post-storm analysis and repairs, all these things that Caitlin’s been talking about.
0:18:38 And obviously, as you mentioned before, data kind of at the key of everything AI related.
0:18:43 Let’s talk about that a little bit. Maybe Caitlin, you can talk about how you and your teams
0:18:49 are training and deploying your AI models. You mentioned before how important it is to show up
0:18:54 at a potential client’s office with pre-trained algorithms kind of ready to go. So maybe talk a
0:18:57 little bit more about your whole approach to using models.
0:19:03 Definitely. We recognized that on day one, it was really important for us to show up with pre-trained
0:19:10 algorithms that were tuned and trained specifically for utility infrastructure. We built our algorithms
0:19:15 from scratch. We spent two years training our algorithms and collecting all of this data,
0:19:21 highly proprietary data from utilities and worked with utilities to understand which anomalies were
0:19:26 highest priority for them so that we could show up on day one with these algorithms that were already
0:19:33 trained. And so we have dozens of different types of pre-trained detections today. We have data sets dating back
0:19:40 over a decade, working with dozens and dozens of utilities across different geographies, both here
0:19:46 in the US and internationally, and also data sets that were collected from a variety of different sources.
0:19:51 As we talked about earlier, data is not standardized in the way it’s collected in this space. And so the
0:19:57 ability for us to train and tune our algorithms with high accuracy meant that we had to work with a variety of
0:20:02 different types of data sets. And so it took us a while to do so. But one thing I wanted to touch on
0:20:08 is Buzz has also started to do a lot more with training our algorithms using synthetic data to be
0:20:15 able to deploy a new algorithm with an anomaly that may not occur with high frequency. So if you think about
0:20:20 some utility issues, they don’t occur with high frequency. For example, smoking in a substation.
0:20:25 You know, we obviously can’t, you know, have a substation fire, light a substation on fire to
0:20:32 train an algorithm. But we still had to train the algorithm to be able to deploy at a substation
0:20:37 and say, hey, we can detect smoke if it occurs. And so we were able to use synthetic data to be able
0:20:43 to train the AI for smoke and substations. We did the same thing for a lot of our animal detections as
0:20:48 well, because capturing enough varied data of animal detections is more challenging. And I want to be,
0:20:54 you know, really clear here to the success of our solution is on a, I guess you’d say the granularity of
0:20:59 number of detections that we can provide was highly because of the amount of real data that we had in order
0:21:05 to be able to leverage synthetic data and use synthetic data successfully, we had to have a lot of real data
0:21:11 in order to be able to use synthetic data more effectively. But there are certain instances where synthetic
0:21:16 data can be quite valuable. And those were just a couple examples of how we’ve been able to use that to
0:21:21 deploy some of those newer algorithms that we have today for our, specifically our power guard solution.
0:21:29 Right. In working across, you know, the United States, but then also globally, are there great differences in the way that
0:21:35 electricity is distributed in different parts of the world and the way that grids are set up? And I mean, my knowledge
0:21:41 is very limited, but I know a little bit about burying power lines underground as opposed to putting them above ground.
0:21:46 So, you know, I’m thinking about it from the data standpoint, but then also just curiosity, like, is the physical
0:21:51 infrastructure generally the same or does it vary greatly from place to place?
0:21:58 We see it varies greatly from place to place, mostly because of when the infrastructure was built, you know, whether it’s
0:21:59 transmission, distribution.
0:22:01 So what’s the difference?
0:22:04 Transmission is high voltage, high voltage infrastructure.
0:22:09 You see those big steel towers more often than not that are traveling far distances.
0:22:10 That’s transmission.
0:22:14 And then distribution are the lower voltage lines, like city lines.
0:22:19 Most of the time you’ll see them are those wooden poles that are throughout cities or surrounding areas.
0:22:24 And so transmissions, much higher voltage, distribution is typically lower voltage.
0:22:29 And so depending on locationally where you are, we’ll see differences in how those towers look.
0:22:35 You know, for example, internationally, we’ve seen a lot more concrete structures than we have here in the U.S.,
0:22:38 whereas in the U.S., we see a lot more steel and wood.
0:22:44 So the way that the structures look can be different internationally versus domestically.
0:22:50 And then, you know, the components deployed on those structures can also look a little bit different depending on the region, where you’re located.
0:22:56 Some infrastructure performs better in different terrain than others, and so that also factors into it.
0:23:01 Right. Kind of going back to the computing side of things, are there particular, and I don’t know if it’s components
0:23:12 or types of disruptions that are harder to detect using, you know, computer vision and AI and everything that just have given you problems for whatever reasons?
0:23:14 Definitely. So maybe I’ll give three different examples.
0:23:20 The first, more granular components, like a small cotter pin that’s missing or loose.
0:23:27 I mean, it’s very hard for an algorithm to tell what’s missing if you don’t have a baseline data set of how it was deployed previously.
0:23:31 So we find that loose or missing can be a little bit more difficult.
0:23:35 So, you know, more granular components can be hard.
0:23:46 We also find that, for example, for something like insulator detections, something as much as the time of day when the image was collected can play into it because shadows can have an impact.
0:23:58 And so when we’re starting to look at different types of damages on an insulator, the time of day that the image was collected and the shadows on that piece of equipment can definitely have an impact on the algorithms.
0:24:00 So something like that can be harder.
0:24:04 So that’s the first thing is granular components or lighting, time of day.
0:24:07 The second is how data was collected.
0:24:16 So depending on distance, angle, resolution, again, lighting, that all can play into the success of how an algorithm is trained as well.
0:24:23 And then the third piece that I will get into is going back to that anomalies that don’t exist with high frequency.
0:24:33 So whether it’s a type of squirrel, a type of snake, or whether it’s, you know, a rusting of a component in a territory where rust may not be prevalent, let’s say.
0:24:33 Right.
0:24:38 That can be harder for us to train and tune the algorithms with high degree of accuracy.
0:24:49 And that’s why we worked with the third piece or the third reason specifically is why we worked with so many utilities from varied regions with tons of different data sets.
0:24:54 We were able to then get more access to anomalies that don’t occur with high frequency.
0:25:07 And maybe one more piece to add to the third talk point here is utilities sometimes have tried to build in-house programs where they’ve tried to train and build algorithms entirely in-house.
0:25:13 And they often come across the same challenge that, you know, there’s not, certain anomalies don’t occur with high frequency.
0:25:18 And so how do you train algorithms without a lot of data of that anomaly being present?
0:25:25 And so that’s where someone like Buzz can come in and help because we do have such a rich data set dating back, you know, so much time working with so many utilities.
0:25:25 Yeah.
0:25:26 Okay.
0:25:27 I asked you what’s challenging.
0:25:30 On the flip side, there are some success stories on your website.
0:25:36 And at the end, we’ll do URLs and such because I know listeners are going to want to learn more about everything you guys are doing.
0:25:41 But maybe there are some success stories that you can quickly highlight, one or two, from some of the work that you’ve been doing.
0:25:42 Yeah, absolutely.
0:25:45 So I’ll highlight one of our recent ones.
0:25:46 It’s with the utility on the East Coast.
0:25:49 And they were doing an insulator replacement program.
0:25:59 They recognized that one of their types of insulators, their porcelain insulators, were failing at a faster rate than a different type of insulators, their polymer insulators.
0:26:08 And this is really common for utilities as different components were more popular and different, I guess you’d say, materials were more popular during different time periods.
0:26:14 Utilities over decades are realizing which materials fail at faster rates than others.
0:26:19 That could be weather-related, that could be capacity-related, could be a number of different reasons.
0:26:20 Sure.
0:26:23 Anyway, they have this insulator replacement program.
0:26:27 And they had this historical helicopter data set.
0:26:31 It was a very, very zoomed out helicopter data set that they had collected.
0:26:39 They were collecting two images per transmission structure, so the high-voltage structures, which is highly zoomed out.
0:26:41 And the insulators were very small in the images.
0:26:43 Where’s Waldo for the insulator?
0:26:46 Yes, not far off from that.
0:26:49 And so these images had the whole structure in it.
0:26:51 So the insulator was quite small.
0:26:57 And we were asked to say, hey, can you do different types of insulator detections?
0:27:03 Could we identify which was a porcelain insulator and which was a polymer insulator so we could help this utility inventory their assets?
0:27:09 We were able to tune our algorithms to get well into the 90% accuracy range within just a couple weeks.
0:27:17 And we were able to take a data set that was 50,000 structures that would have taken them months to analyze this data set.
0:27:25 We were able to do so in a matter of hours once our algorithms were tuned and trained with a high degree of accuracy and a high degree of success.
0:27:34 And what this does for utilities is this helps them, one, save mass amounts of time because they didn’t have to have someone manually analyzing all of this data.
0:27:37 They could just take the action on the data set that was analyzed.
0:27:48 But the second thing is by Buzz being able to work with this highly zoomed out data, we were able to save them mass amount of time from going out in the field and recollecting all this different inspection data.
0:27:51 So two folds, time savings and cost savings.
0:27:55 And we could help them then prioritize where they should go through the replacement of those insulators.
0:28:01 And one thing I want to mention here is, you know, time savings and cost savings are huge for us here at Buzz.
0:28:06 But I want to spend a second talking about the workforce enablement because workforce enablement is huge.
0:28:11 You know, when we started Buzz seven, eight years ago, there was a lot of concern around,
0:28:13 Hey, is AI going to take the person’s job?
0:28:16 And I want to be really clear and say,
0:28:18 Thankfully, all that concern is dissipated.
0:28:19 Nobody asked that question anymore.
0:28:25 You know, we still, we still hear that question, you know, in this, in this type of an industry.
0:28:30 But new thing, you know, there’s so much new and people adjust and it takes time.
0:28:31 And yeah, sure, sure.
0:28:36 And I think what we’ve, you know, realized is as more data has been collected, as more inspections are happening,
0:28:43 the utility, highly trained field workers, engineers, personnel, their job is to take action on the data,
0:28:49 to make the reason-based decisions so they can go out and optimize their maintenance program.
0:28:54 They shouldn’t be spending their time six to eight months manually looking through images.
0:29:04 And so something like our solution allowed these field workers, these engineers to be able to go out and fast track the timeline for that program,
0:29:07 as opposed to spending so much time manually analyzing data.
0:29:09 So I just wanted to take a second to touch on that, too.
0:29:10 No, absolutely.
0:29:11 It’s a great point.
0:29:17 And I think you said it very well that, you know, let the machines do what they’re good at,
0:29:21 to give the human what the human can then go use to do what they’re really good at.
0:29:25 And that’s, you know, and I think as you were speaking, I was thinking about, you know,
0:29:28 content creation, my industry, but other industries.
0:29:30 And like, yeah, that might work as a rule of thumb.
0:29:31 I like that.
0:29:32 So, no, well said.
0:29:33 And it’s hugely important.
0:29:35 So I’m not going to hold you to this.
0:29:38 I’m not going to ask you to speak, you know, outside of your comfort zone.
0:29:42 But as you think ahead, the next five years, seven to ten years, maybe it’s less.
0:29:49 How do you see AI continuing to make an impact on, you know, and I’ll say the energy sector,
0:29:55 but you can go as narrow or as wide as you like, you know, based on what Buzz does and doesn’t do
0:29:56 and all of your experiences.
0:29:58 And, you know, where’s this all headed?
0:30:01 And what are the impacts AI is going to make?
0:30:03 That is one small question.
0:30:04 One small.
0:30:05 We like to end on the small ones.
0:30:08 I actually do have a very small one for you after this, but we’ll give you the hard
0:30:09 one first.
0:30:11 Yeah, it’s a really interesting question.
0:30:18 We are just at the tip of the iceberg of seeing AI enter into the energy sector and start to
0:30:20 provide real value in the energy sector.
0:30:25 And I say that and I find it interesting because we were an AI company that launched eight years
0:30:27 ago before AI was cool.
0:30:28 Yeah, I know.
0:30:30 You’re old veterans here, yeah.
0:30:30 We are.
0:30:31 We are.
0:30:34 We’re ancient in the terms of AI in this industry.
0:30:40 But it’s really interesting because we see that we certainly here at Buzz are just starting
0:30:45 to touch the tip of the iceberg of what’s possible from an asset inventorying standpoint to an
0:30:50 optimized maintenance standpoint to a better data-driven decision-making standpoint.
0:30:55 But utilities are just at the beginning of even their data collection journey in a lot
0:30:55 of ways.
0:31:01 So whether it’s inspections, whether it’s demand forecasting, load growth, there’s so
0:31:07 many different ways that AI is helping provide better data-driven decision-making in the
0:31:08 energy space.
0:31:11 And I think there’s just so much room for huge value.
0:31:16 So as it pertains to the next five to 10 years, I really think we’re setting ourselves up for a
0:31:23 lot of exciting ways that innovation can help come in and take over some of these more, I guess,
0:31:29 you know, historical or mundane processes and help the workforce, help the crews make better
0:31:30 data-driven decisions.
0:31:34 So that’s what I’m most excited about, I would say, in the next five to 10 years.
0:31:41 This industry, specifically the utility world, has been very reactive to problems that have
0:31:44 occurred as opposed to being proactive and preventative.
0:31:51 And I’m excited for us to be able to start taking steps forward into a more proactive and
0:31:56 preventative space so that utilities can operate in a, I guess you should say, more data-driven
0:32:00 way as opposed to being so reactionary to everything that’s happening.
0:32:05 And as you look at things like data centers coming online, electrification, you look at renewables,
0:32:11 we have all these massive initiatives that are coming at us at rapid pace, you know, rapid speed.
0:32:15 And utilities, as the, really the backbone of our infrastructure, are trying to figure out
0:32:17 how they can respond to it.
0:32:20 As the backbone of our economy, you know, they’re trying to figure out how they can respond to
0:32:24 all these different initiatives that are happening at such quick rates.
0:32:31 And you look at the age of our utility infrastructure, some of it being over 100 years old, it wasn’t
0:32:37 set up for a lot of these new initiatives like renewables coming onto the grid with bi-directional
0:32:38 passing of energy.
0:32:44 So as you look at the ways that AI can come in and help, it helps be able to leverage mass
0:32:49 amounts of data, turn that into better decisions, and help us as we’re looking to upgrade and
0:32:51 modernize our infrastructure more broadly.
0:32:56 You know, for unfairly throwing you a big answer with a, or a big question, rather, with
0:33:00 a five to 10-year time span, when I heard it say back to me, I was like, we can’t ask her
0:33:00 that.
0:33:01 So much is going to change.
0:33:02 That was a fantastic answer.
0:33:03 So thank you.
0:33:03 All right.
0:33:05 Last question for you.
0:33:06 A little change of pace.
0:33:06 This is an easier one.
0:33:12 Are there AI tools that you are currently really using, enjoying on a regular basis,
0:33:14 whether for work or personal life?
0:33:18 And you don’t have to get into what you’re using them for, but any tools that, you know,
0:33:21 have made your day-to-day a little bit better?
0:33:22 Oh, that’s a great question.
0:33:29 I would say from a, you know, as a company of our size, we are constantly looking, we’re
0:33:30 about 35 people.
0:33:31 Okay.
0:33:31 Yeah.
0:33:35 So as a company of our size, you know, we use a variety of different tools to help us
0:33:38 as we’re looking to optimize and be more efficient.
0:33:41 I would say I’m using AI in several different ways of work.
0:33:47 But one thing that I find incredibly valuable is using tools like AI to help me sift through
0:33:47 my inbox.
0:33:54 I know that’s kind of a funny thing, but if you can get a couple hours back and productivity,
0:33:56 make sure you don’t miss those important emails.
0:34:00 I’m a zero inbox kind of person, but I find it incredibly valuable.
0:34:06 You’ve got a tool hooked up to your inbox now and it just, before you look at it, goes
0:34:06 through.
0:34:09 And does it give you a summary or sort of flag certain messages?
0:34:11 It flags them and helps prioritize.
0:34:12 Yeah.
0:34:14 Prioritize which ones need immediate responses.
0:34:19 And I find, I find not to be incredibly valuable, but you know, as our team here at Buzz, we’re
0:34:25 looking to use AI to help make our process as efficient as possible as we’re an AI company
0:34:27 delivering an AI solution to the market.
0:34:31 Our team is, especially our engineering team is very excited about the ways we can use things
0:34:35 like generative AI and all of that to help improve our work here at Buzz.
0:34:35 Awesome.
0:34:37 Caitlin, this has been a pleasure.
0:34:38 Thank you so much for taking the time.
0:34:39 I learned a ton.
0:34:45 And for listeners who want to find out more about all the work Buzz is doing and the ways
0:34:50 that machine learning and AI are impacting how we get our electricity, places online they
0:34:53 can go, company website, social media, where should they head?
0:34:53 Yes.
0:34:56 You can find us at buzzsolutions.com.
0:34:57 That’s our website.
0:34:58 And then we’re also very active on LinkedIn.
0:35:01 So you can check out the Buzz Solutions LinkedIn page.
0:35:06 We have some great examples of videos of some of our detections, ways in which utilities have
0:35:10 deployed AI, how you can use AI in the utility inspection workflow.
0:35:14 So if you’re interested in learning more, I would definitely check both of those out.
0:35:14 Excellent.
0:35:15 Well, again, thank you.
0:35:21 And best of luck to you and your teams for everything you’re doing, you know, as a person
0:35:23 who relies on electricity every day.
0:35:24 I’m rooting for you.
0:35:25 Thank you.
0:35:26 Thank you so much for having me.
0:35:28 I really enjoyed the conversation today.
0:35:58 Thank you.
0:36:17 Thank you.
0:00:20 Buzz Solutions is on a mission to enhance the safety and efficiency of the electric grid
0:00:25 through innovative data analytics and machine learning. A member of NVIDIA’s inception program
0:00:29 for startups, Buzz’s visual intelligence empowers utility companies to better monitor
0:00:34 and manage their infrastructure. The result is improved reliability and reduced outages.
0:00:39 Here to explain how Buzz does it and to talk about the impact AI can have on making our electric
0:00:46 grids safer and more robust is Caitlin Albertoli, CEO and co-founder of Buzz Solutions. Caitlin,
0:00:51 welcome, and thanks so much for joining the NVIDIA AI podcast. Thanks so much for having me, Noah. I’m
0:00:56 excited to be here. So we’re two Californians, and I mean, I think it’s everywhere now, but certainly in
0:01:01 our state, the electric grid has been a source of news for quite some time now. It’s something
0:01:06 that’s on lots of people’s minds. So I’m very excited to learn more about what Buzz is doing
0:01:12 and the impact that AI can have going forward to really help all of us because we all use electricity,
0:01:18 right? So maybe we can get into it with a little bit about your background and how you came to co-found
0:01:24 Buzz Solutions. Absolutely. Thanks for the question. Happy to share. So we launched Buzz Solutions from a
0:01:32 launchpad course at Stanford University in the spring of 2017. My background is actually not from the electric and
0:01:38 power industry. Prior to Buzz, I was working in finance, and I also ran a nonprofit in the sustainable food
0:01:45 world. Before that, I grew up in Southern California in a town that was close to the nuclear power plant, the
0:01:51 San Onofre power plant. So that was really my first exposure to the power world, the power industry. It was
0:01:57 actually shut down during the time that I was living in Southern California. So that was an interesting dynamic
0:02:02 to, you know, grow up with and see. Yeah. Did that leave an impression on you? I mean, did it, did you think more
0:02:07 about where power comes from than perhaps, you know, the average kid? Definitely. It definitely got me thinking
0:02:14 more about our different generation sources and renewable energy more broadly. And so when I went into Stanford, I was
0:02:19 certainly looking for more opportunities to learn about about generation and power as a whole.
0:02:24 That’s great. So there’s a lot of steps on the way, as you mentioned, with, you know, finance and sustainable
0:02:28 food. But how did you wind up co-founding Buzz? How did Buzz come to be?
0:02:35 Sure. So my co-founder, Vic, and I met in this launchpad course that was in the School of Civil and Environmental
0:02:40 Engineering. It was about building a startup in the School of Civil and Environmental Engineering focused
0:02:46 on energy and sustainability. And there was something that called me to this class because I was always
0:02:52 really entrepreneurial and interested in learning more about starting a business. But the, I guess,
0:02:57 the intersection of energy and sustainability, given my background, with entrepreneurship seems like
0:03:03 a really unique opportunity. And so that was the course that, that we ultimately met. And when we
0:03:09 were originally exploring ideas for, you know, what we wanted to, I should say, not really found a company
0:03:15 at that time, but what we wanted to use for this class, we were looking at the wind turbine market. We were
0:03:21 looking at all that was happening with optimization of placement of wind turbines and how you could use
0:03:26 drones and technology to understand where it was best to place those wind turbines. But in those
0:03:33 conversations with some of those, you know, those wind turbine companies and manufacturers, we very
0:03:37 quickly were tipped off to the power space. You know, everyone said, this is a huge market and
0:03:43 there’s definitely an opportunity in wind. But have you seen what’s happening with inspections in the
0:03:50 power space? And kind of an interesting time, because if you think about 2017, that was before a lot of the
0:03:55 big wildfires, before a lot of the power outages, and before some of the major blackouts that, of course, we know
0:04:00 about today and what has unfortunately transpired since that time. But we ended up taking that
0:04:06 conversation and leverage the Stanford alumni network to interview 35 investor on utilities. We were able
0:04:13 to get conversations with 35 investor on utilities. Can you imagine that that many? And they all had the
0:04:19 same story. They were collecting a lot more data, or they had plans to collect a lot more inspection data,
0:04:24 whether it was due to, you know, improve technology that allowed them to do more inspections, and also
0:04:29 regulations that were driving more inspections. But they didn’t have a way to manage or analyze that
0:04:33 data. And so that was really kind of the genesis of ultimately how we launched Buzz.
0:04:38 Yeah. So Caitlin, it’s funny, because I’m thinking about that kind of pre-pandemic time and going to
0:04:46 NVIDIA GTC back, you know, probably around 2018, 2019, maybe 2017, and watching a demo of drones and
0:04:52 kind of a, you know, is it a demo in the convention space, but a drone in a simulated warehouse environment
0:04:57 flying around doing inspections and doing what’s it called pick and pack and that kind of thing.
0:05:02 So what you’re talking about rings a bell for me, if in a different space. But how did you, you know,
0:05:06 when you’re talking about there’s all this data, what do we do with it? And to me, that says AI,
0:05:11 but this is, you know, seven, eight years later. So back then, did you see instantly, did you see
0:05:15 all the data and think, oh, machine learning, AI, we can do something with this? Or kind of how did
0:05:20 that leap to AI happen? It’s a really good question, because at that time, for a lot of utilities,
0:05:26 it was still just an idea, or it was still just a plan that they were looking to put in place. You know,
0:05:31 they were buying their first drones, they were training their first pilots. And on top of that,
0:05:36 they were still doing, and present day still are doing, a ton with helicopters. So they were
0:05:41 starting to actually collect images with helicopters at that time, while also building drone programs.
0:05:47 And so even in our first days of Buzz, a lot of our conversations with utilities revolved around,
0:05:52 okay, we’re starting to use drones, we’re starting to collect data with drones. How do we scale this
0:05:58 program? How do we collect data in the best way that’s going to make AI the most successful,
0:06:03 I guess you would say? And we still have a lot of those conversations today. But data wasn’t
0:06:08 standardized in that space, and it still isn’t standardized. But if you can think in 2017,
0:06:14 that a more traditional industry like utilities, where they had five to seven year plans to scale
0:06:19 these massive drone programs, I mean, it’s lightning speed for some of these more traditional industries
0:06:25 like utilities. So it was more an idea at that time, but they had all of the right, I guess you would say,
0:06:29 pieces to the puzzle to be able to build some of these programs.
0:06:34 And so what’s happened along the way, and I want to get into, you know, some of the use cases,
0:06:39 some of the challenges and problems that you’ve solved and are helping energy companies solve,
0:06:44 but I don’t want to skip over the growth in between. Walk us through that. And if there are some of those,
0:06:49 you know, use cases and things that kind of popped up and are still prevalent, definitely dig into those if you would.
0:06:56 Sure. It was an interesting time back in the 2017-2018 timeframe. We’re talking about these concepts like
0:07:01 artificial intelligence, AI, and a lot of our user base at that time, they weren’t really actively using
0:07:06 software platforms. A lot of them were still manually drafting inspection reports. So when you’re talking
0:07:10 about going from manually drafting inspection reports to using AI…
0:07:12 And these are the utility companies you’re talking about.
0:07:14 Correct. Yeah. These are the utility companies.
0:07:16 Yeah. Digital transformation. Yeah.
0:07:23 Exactly. Lineman, field technicians, engineers, a lot of them were still manually drafting inspection
0:07:28 reports. And some of our, you know, lineman and field technicians that were using our solution,
0:07:32 we’re having to go through, you know, thousands of images, tens of thousands and hundreds of thousands
0:07:38 of images manually before they’re looking to use a tool like this. And so the reason I give you that
0:07:45 background is in order to show up to a utility and ultimately, you know, bring value with a solution
0:07:51 like ours, we had to plug into the utility workflow. And so from day one, when we were working with
0:07:56 utilities to build this solution, to train our algorithms, to build our platform, we were really
0:08:02 utility centric. We wanted to make sure that the algorithms that we were training were very closely
0:08:07 aligned with how a utility subject matter expert would have labeled this data manually.
0:08:14 Is that, or was that back then sort of standardized or was it that, you know, from company to company,
0:08:20 even within a company from sort of inspector to inspector, the format of those reports was just
0:08:22 sort of a free form, I guess.
0:08:27 I would say there’s, there’s definitely a lot of subjectivity as it pertains to how, how lineman or
0:08:33 field technicians would label the data and how they would classify a significance, as we call them,
0:08:39 like anomaly or should say a physical defect on a component is how someone could think of it more
0:08:44 commonly. But it’s, if you can see it, the defect on a component or the anomaly, that could be a little
0:08:50 subjective of how someone classifies it. But we see that there’s a pretty standard set of labels or total
0:08:56 classifications that a utility is looking for. And we worked with research organizations like EPRI,
0:09:01 which is the Electric Power Research Institute, one of the leading electric research institutes,
0:09:05 along with utilities to make sure that we got that nomenclature correct.
0:09:12 And when you’re talking about defects, anomalies, what kinds of things are, you know, are you looking
0:09:16 for? Are they looking for? I, I not knowing much about it all. I’m just imagining, you know,
0:09:21 there’s a squirrel chewing through a line or there’s a giant explosion. What kinds of things are being,
0:09:24 you know, captured as anomalies? Actually, squirrels do have a big impact.
0:09:30 So those are the ones I always hear stories about, you know, like, oh, what was the outage last? Oh,
0:09:34 is this squirrel at the substation again? Or, but is that really like they’re that one? Yeah, it is.
0:09:42 Actually in substations in particular, a substation is where a utility changes the voltage. So it goes from,
0:09:48 from highly energized lines to lower energized lines or vice versa. So the substation is the facility where
0:09:53 that that change takes place. A substation, a lot of our questions that we get from utilities are,
0:10:01 hey, can you do things like birds, squirrels, snakes? Snakes is a huge one. Coyotes, mountain lions. So
0:10:07 we actually get a lot of animal requests as well, because animals, if you think about it, you know,
0:10:11 they’re sitting on the infrastructure or they’re chewing through something. It could have a significant
0:10:17 impact on the equipment itself. And a person may not be out there to be able to see that,
0:10:22 but they may start to get a sensor that’s tripping. And they may say, oh, there’s,
0:10:27 there’s something wrong here, but they may not know what caused it. And so having the ability to
0:10:31 leverage visual data to say, oh, okay, there was a squirrel that came in contact with this or a snake
0:10:35 can be really helpful for that utility to make a better decision on maintenance then.
0:10:40 Just to get a sense of the scale that we’re talking about that you’re dealing with, the Buzz is dealing
0:10:47 with, what’s the best way to, like, how much data, how many square miles, how many, you know, data points
0:10:52 or images? How do you, what’s the best way to express, you know, the scale of what these companies
0:10:54 are doing and what Buzz is doing to help them?
0:10:59 Sure. Some of our utility customers are collecting hundreds of thousands and millions of images a
0:11:05 year. And that’s just as a part of their standard inspection processes. Utilities have various
0:11:10 different types of inspections. So they could be doing what’s called a flyby inspection, which is
0:11:15 just looking for anything major glaring that’s happening along the infrastructure. Or they could
0:11:20 be getting down to a comprehensive inspection where they have a drone that’s flying and taking multiple
0:11:27 pictures to look for things like a insulator disc that’s broken or, or a crack in a cross arm or a
0:11:32 woodpecker hole in a cross arm, or even something as small as a cotter pin that could be missing or
0:11:37 loose, you know, a bolt or something like that, that’s holding the insulators to the line. So there’s all
0:11:42 these different types of inspections that take place throughout the year. And a large utility may be
0:11:47 collecting millions of images of their infrastructure that they’re having to sift through on an annual
0:11:55 basis. We see that utilities are doing data collection with helicopters, drones, big swing aircrafts, ground
0:11:59 based data collection, some are using satellites. So there’s a variety of different ways that they’re
0:12:03 collecting this data. And typically, it’ll serve a specific purpose for that utility.
0:12:09 Got it. Okay. And then, so how does, how does buzz work? You have a couple of different products listed on
0:12:14 your website, I would imagine. Well, I don’t know if that’s the full scope of it or not. But maybe walk us through
0:12:19 some of the solutions that, you know, your customers, the utility companies are employing and kind of how
0:12:24 it works from, you know, imagining data collection and analysis and action, but that may be entirely
0:12:30 wrong. So how does it work? Sure. Great question. So at Buzz, we are here to modernize energy infrastructure
0:12:37 with actionable intelligence. The central focus of what we do here is to help utilities take mass amounts
0:12:43 of raw data and use AI, machine learning, computer vision to turn it into actionable information or
0:12:48 non-actionable information if it’s not significant enough for utility to actually take action on it
0:12:54 right away. And what that means for us is we’re taking in all these different visual data sources.
0:12:59 So as you’ll see on our website, we have two different products that are listed. Our first one is more
0:13:04 inspection oriented. It’s called Power AI. And so that’s where we’re taking in all of this inspection
0:13:10 data that I’ve been talking about from the drones or helicopters, ground-based. It’s really linear
0:13:15 infrastructure data that we help the utility by managing it, meaning we’ll ingest all that data.
0:13:21 They can upload it to our platform. We’ll help them map it so that they know exactly where all the images
0:13:26 are on their corridor. And then we’ll analyze it with our machine learning, computer vision algorithms.
0:13:34 What was really important for us as we were training our algorithms and before we ultimately deployed was
0:13:39 that we could show up to a utility with pre-trained algorithms, meaning on day one, we could start
0:13:45 analyzing their data and start delivering immediate value. So our Power AI platform has pre-trained
0:13:50 algorithms. We’ll analyze all that data. And for us, we can analyze an image in a fraction of a second.
0:13:56 So it’s really quick, really fast. And then once those images are analyzed, we display them on a
0:14:03 dashboard for a utility to be able to go through and they can actually edit, adjust labels, add,
0:14:08 delete labels as they see fit. And then all of that gets pushed into an inspection report. We can deploy
0:14:14 that in their workflow a variety of ways, whether it’s directly into a work management system or to a
0:14:20 geospatial information system, which is a GIS system for mapping and managing their inventory.
0:14:25 We can deploy it in a variety of different ways for their workflow. So that’s Power AI. And then our
0:14:32 second product is called Power Guard. And that’s the one I mentioned kind of as we do animal detections
0:14:38 and all that type of thing. But Power Guard is a real-time or a near real-time alerting system
0:14:44 that deploys on a fixed camera. And so for Power Guard, we’re doing things, same component monitoring
0:14:50 that we’re doing with Power AI. But here we’re doing it as an alerting system in a substation or
0:14:56 in a facility where a utility needs more of those kind of eyes and continuous monitoring. So we call
0:15:02 Power Guard our continuous monitoring solution for security at a substation, whether that’s an intruder,
0:15:10 person, car, animal, squirrel, exactly, whether it’s safety. So if someone’s injured on the job
0:15:14 or they’re injured in their work, and then we do all sorts of component monitoring, whether that’s
0:15:19 overheating of components or physical degradation on those components that can be seen. And we send
0:15:24 those alerts 24-7 to the relevant teams at the utility. So that’s kind of our two different products.
0:15:26 It’s all vision-based as well.
0:15:32 Right, of course. So in a, I don’t know, like an emergency situation, a wildfire, a storm,
0:15:38 or something else that’s causing, you know, whether it’s a sudden unexpected blackout, or if a utility
0:15:44 company decides they need to roll out, you know, brownout blackouts, I can understand how Power Guard
0:15:49 obviously would alert, you know, there’s something’s gone wrong at this substation. But how do Buzz’s solutions
0:15:54 play into helping utility companies deal with, you know, these kind of real-time
0:15:58 emergency situations that are coming up?
0:16:03 So for things like Power Guard, we also do smoke detections. So we can detect smoke or overheating
0:16:04 or sparking issues.
0:16:05 Is that all visual?
0:16:10 Yeah. So we’re able to alert to any of those incidences that are happening, whether it’s within
0:16:16 the substation or at the utility facility, or even in the surrounding area. So we can alert to
0:16:21 any potential fire risk that’s happening there. On the Power AI side for inspections,
0:16:27 we’re identifying critical risk issues that could cause a wildfire before they start.
0:16:32 It’s preventative maintenance, basically. So we’re identifying, you know, for example,
0:16:36 rusted components that could ultimately rust out and cause a line to drop down.
0:16:43 We are able to do those types of detections on a transmission tower distribution pole. So a high
0:16:49 voltage tower or a low voltage city line, we’re able to do those detections. And then we also can
0:16:54 do sparking detections as well. So if there’s any sort of a physical component that could cause a
0:16:59 sparking issue, we’re able to detect that on the physical components. And then we do things like
0:17:04 vegetation encroachment detections. So we can detect where vegetation is encroaching on a line.
0:17:10 And where it’s particularly valuable for utility is we can say, hey, there’s vegetation encroaching,
0:17:16 let’s say on this distribution or a lower voltage city line. There’s vegetation encroaching on this
0:17:22 distribution pole. And you also have five different anomalies or defects that are on this particular
0:17:27 tower, this particular pole. This is a high risk pole that you should definitely be maintaining. So we do
0:17:33 those types of things to help with wildfire risk mitigation. And then on the storm side of things,
0:17:40 utilities are starting to deploy more vision-based inspections like drone technologies post-storm
0:17:47 to go out and fly corridors so they can understand what sorts of anomalies exist on the infrastructure
0:17:53 before they re-energize that line. And by using Buzz, they’re able to analyze and process through that
0:17:57 data really quickly. So they understand exactly what they need to replace prior to re-energizing.
0:18:03 So we’re able to save mass amounts of time and help direct the field crews where to go to upgrade
0:18:05 and restore that infrastructure post-storm.
0:18:12 That’s fantastic. I’m speaking with Caitlin Albertoli. Caitlin is CEO and co-founder of Buzz Solutions.
0:18:17 And we’ve been talking about Caitlin’s journey to starting Buzz, but Buzz’s journey in their work
0:18:23 using machine learning, using computer vision, other AI technologies to help electric grids,
0:18:29 help utility companies deliver power safely, more effectively, do preventative maintenance,
0:18:33 post-storm analysis and repairs, all these things that Caitlin’s been talking about.
0:18:38 And obviously, as you mentioned before, data kind of at the key of everything AI related.
0:18:43 Let’s talk about that a little bit. Maybe Caitlin, you can talk about how you and your teams
0:18:49 are training and deploying your AI models. You mentioned before how important it is to show up
0:18:54 at a potential client’s office with pre-trained algorithms kind of ready to go. So maybe talk a
0:18:57 little bit more about your whole approach to using models.
0:19:03 Definitely. We recognized that on day one, it was really important for us to show up with pre-trained
0:19:10 algorithms that were tuned and trained specifically for utility infrastructure. We built our algorithms
0:19:15 from scratch. We spent two years training our algorithms and collecting all of this data,
0:19:21 highly proprietary data from utilities and worked with utilities to understand which anomalies were
0:19:26 highest priority for them so that we could show up on day one with these algorithms that were already
0:19:33 trained. And so we have dozens of different types of pre-trained detections today. We have data sets dating back
0:19:40 over a decade, working with dozens and dozens of utilities across different geographies, both here
0:19:46 in the US and internationally, and also data sets that were collected from a variety of different sources.
0:19:51 As we talked about earlier, data is not standardized in the way it’s collected in this space. And so the
0:19:57 ability for us to train and tune our algorithms with high accuracy meant that we had to work with a variety of
0:20:02 different types of data sets. And so it took us a while to do so. But one thing I wanted to touch on
0:20:08 is Buzz has also started to do a lot more with training our algorithms using synthetic data to be
0:20:15 able to deploy a new algorithm with an anomaly that may not occur with high frequency. So if you think about
0:20:20 some utility issues, they don’t occur with high frequency. For example, smoking in a substation.
0:20:25 You know, we obviously can’t, you know, have a substation fire, light a substation on fire to
0:20:32 train an algorithm. But we still had to train the algorithm to be able to deploy at a substation
0:20:37 and say, hey, we can detect smoke if it occurs. And so we were able to use synthetic data to be able
0:20:43 to train the AI for smoke and substations. We did the same thing for a lot of our animal detections as
0:20:48 well, because capturing enough varied data of animal detections is more challenging. And I want to be,
0:20:54 you know, really clear here to the success of our solution is on a, I guess you’d say the granularity of
0:20:59 number of detections that we can provide was highly because of the amount of real data that we had in order
0:21:05 to be able to leverage synthetic data and use synthetic data successfully, we had to have a lot of real data
0:21:11 in order to be able to use synthetic data more effectively. But there are certain instances where synthetic
0:21:16 data can be quite valuable. And those were just a couple examples of how we’ve been able to use that to
0:21:21 deploy some of those newer algorithms that we have today for our, specifically our power guard solution.
0:21:29 Right. In working across, you know, the United States, but then also globally, are there great differences in the way that
0:21:35 electricity is distributed in different parts of the world and the way that grids are set up? And I mean, my knowledge
0:21:41 is very limited, but I know a little bit about burying power lines underground as opposed to putting them above ground.
0:21:46 So, you know, I’m thinking about it from the data standpoint, but then also just curiosity, like, is the physical
0:21:51 infrastructure generally the same or does it vary greatly from place to place?
0:21:58 We see it varies greatly from place to place, mostly because of when the infrastructure was built, you know, whether it’s
0:21:59 transmission, distribution.
0:22:01 So what’s the difference?
0:22:04 Transmission is high voltage, high voltage infrastructure.
0:22:09 You see those big steel towers more often than not that are traveling far distances.
0:22:10 That’s transmission.
0:22:14 And then distribution are the lower voltage lines, like city lines.
0:22:19 Most of the time you’ll see them are those wooden poles that are throughout cities or surrounding areas.
0:22:24 And so transmissions, much higher voltage, distribution is typically lower voltage.
0:22:29 And so depending on locationally where you are, we’ll see differences in how those towers look.
0:22:35 You know, for example, internationally, we’ve seen a lot more concrete structures than we have here in the U.S.,
0:22:38 whereas in the U.S., we see a lot more steel and wood.
0:22:44 So the way that the structures look can be different internationally versus domestically.
0:22:50 And then, you know, the components deployed on those structures can also look a little bit different depending on the region, where you’re located.
0:22:56 Some infrastructure performs better in different terrain than others, and so that also factors into it.
0:23:01 Right. Kind of going back to the computing side of things, are there particular, and I don’t know if it’s components
0:23:12 or types of disruptions that are harder to detect using, you know, computer vision and AI and everything that just have given you problems for whatever reasons?
0:23:14 Definitely. So maybe I’ll give three different examples.
0:23:20 The first, more granular components, like a small cotter pin that’s missing or loose.
0:23:27 I mean, it’s very hard for an algorithm to tell what’s missing if you don’t have a baseline data set of how it was deployed previously.
0:23:31 So we find that loose or missing can be a little bit more difficult.
0:23:35 So, you know, more granular components can be hard.
0:23:46 We also find that, for example, for something like insulator detections, something as much as the time of day when the image was collected can play into it because shadows can have an impact.
0:23:58 And so when we’re starting to look at different types of damages on an insulator, the time of day that the image was collected and the shadows on that piece of equipment can definitely have an impact on the algorithms.
0:24:00 So something like that can be harder.
0:24:04 So that’s the first thing is granular components or lighting, time of day.
0:24:07 The second is how data was collected.
0:24:16 So depending on distance, angle, resolution, again, lighting, that all can play into the success of how an algorithm is trained as well.
0:24:23 And then the third piece that I will get into is going back to that anomalies that don’t exist with high frequency.
0:24:33 So whether it’s a type of squirrel, a type of snake, or whether it’s, you know, a rusting of a component in a territory where rust may not be prevalent, let’s say.
0:24:33 Right.
0:24:38 That can be harder for us to train and tune the algorithms with high degree of accuracy.
0:24:49 And that’s why we worked with the third piece or the third reason specifically is why we worked with so many utilities from varied regions with tons of different data sets.
0:24:54 We were able to then get more access to anomalies that don’t occur with high frequency.
0:25:07 And maybe one more piece to add to the third talk point here is utilities sometimes have tried to build in-house programs where they’ve tried to train and build algorithms entirely in-house.
0:25:13 And they often come across the same challenge that, you know, there’s not, certain anomalies don’t occur with high frequency.
0:25:18 And so how do you train algorithms without a lot of data of that anomaly being present?
0:25:25 And so that’s where someone like Buzz can come in and help because we do have such a rich data set dating back, you know, so much time working with so many utilities.
0:25:25 Yeah.
0:25:26 Okay.
0:25:27 I asked you what’s challenging.
0:25:30 On the flip side, there are some success stories on your website.
0:25:36 And at the end, we’ll do URLs and such because I know listeners are going to want to learn more about everything you guys are doing.
0:25:41 But maybe there are some success stories that you can quickly highlight, one or two, from some of the work that you’ve been doing.
0:25:42 Yeah, absolutely.
0:25:45 So I’ll highlight one of our recent ones.
0:25:46 It’s with the utility on the East Coast.
0:25:49 And they were doing an insulator replacement program.
0:25:59 They recognized that one of their types of insulators, their porcelain insulators, were failing at a faster rate than a different type of insulators, their polymer insulators.
0:26:08 And this is really common for utilities as different components were more popular and different, I guess you’d say, materials were more popular during different time periods.
0:26:14 Utilities over decades are realizing which materials fail at faster rates than others.
0:26:19 That could be weather-related, that could be capacity-related, could be a number of different reasons.
0:26:20 Sure.
0:26:23 Anyway, they have this insulator replacement program.
0:26:27 And they had this historical helicopter data set.
0:26:31 It was a very, very zoomed out helicopter data set that they had collected.
0:26:39 They were collecting two images per transmission structure, so the high-voltage structures, which is highly zoomed out.
0:26:41 And the insulators were very small in the images.
0:26:43 Where’s Waldo for the insulator?
0:26:46 Yes, not far off from that.
0:26:49 And so these images had the whole structure in it.
0:26:51 So the insulator was quite small.
0:26:57 And we were asked to say, hey, can you do different types of insulator detections?
0:27:03 Could we identify which was a porcelain insulator and which was a polymer insulator so we could help this utility inventory their assets?
0:27:09 We were able to tune our algorithms to get well into the 90% accuracy range within just a couple weeks.
0:27:17 And we were able to take a data set that was 50,000 structures that would have taken them months to analyze this data set.
0:27:25 We were able to do so in a matter of hours once our algorithms were tuned and trained with a high degree of accuracy and a high degree of success.
0:27:34 And what this does for utilities is this helps them, one, save mass amounts of time because they didn’t have to have someone manually analyzing all of this data.
0:27:37 They could just take the action on the data set that was analyzed.
0:27:48 But the second thing is by Buzz being able to work with this highly zoomed out data, we were able to save them mass amount of time from going out in the field and recollecting all this different inspection data.
0:27:51 So two folds, time savings and cost savings.
0:27:55 And we could help them then prioritize where they should go through the replacement of those insulators.
0:28:01 And one thing I want to mention here is, you know, time savings and cost savings are huge for us here at Buzz.
0:28:06 But I want to spend a second talking about the workforce enablement because workforce enablement is huge.
0:28:11 You know, when we started Buzz seven, eight years ago, there was a lot of concern around,
0:28:13 Hey, is AI going to take the person’s job?
0:28:16 And I want to be really clear and say,
0:28:18 Thankfully, all that concern is dissipated.
0:28:19 Nobody asked that question anymore.
0:28:25 You know, we still, we still hear that question, you know, in this, in this type of an industry.
0:28:30 But new thing, you know, there’s so much new and people adjust and it takes time.
0:28:31 And yeah, sure, sure.
0:28:36 And I think what we’ve, you know, realized is as more data has been collected, as more inspections are happening,
0:28:43 the utility, highly trained field workers, engineers, personnel, their job is to take action on the data,
0:28:49 to make the reason-based decisions so they can go out and optimize their maintenance program.
0:28:54 They shouldn’t be spending their time six to eight months manually looking through images.
0:29:04 And so something like our solution allowed these field workers, these engineers to be able to go out and fast track the timeline for that program,
0:29:07 as opposed to spending so much time manually analyzing data.
0:29:09 So I just wanted to take a second to touch on that, too.
0:29:10 No, absolutely.
0:29:11 It’s a great point.
0:29:17 And I think you said it very well that, you know, let the machines do what they’re good at,
0:29:21 to give the human what the human can then go use to do what they’re really good at.
0:29:25 And that’s, you know, and I think as you were speaking, I was thinking about, you know,
0:29:28 content creation, my industry, but other industries.
0:29:30 And like, yeah, that might work as a rule of thumb.
0:29:31 I like that.
0:29:32 So, no, well said.
0:29:33 And it’s hugely important.
0:29:35 So I’m not going to hold you to this.
0:29:38 I’m not going to ask you to speak, you know, outside of your comfort zone.
0:29:42 But as you think ahead, the next five years, seven to ten years, maybe it’s less.
0:29:49 How do you see AI continuing to make an impact on, you know, and I’ll say the energy sector,
0:29:55 but you can go as narrow or as wide as you like, you know, based on what Buzz does and doesn’t do
0:29:56 and all of your experiences.
0:29:58 And, you know, where’s this all headed?
0:30:01 And what are the impacts AI is going to make?
0:30:03 That is one small question.
0:30:04 One small.
0:30:05 We like to end on the small ones.
0:30:08 I actually do have a very small one for you after this, but we’ll give you the hard
0:30:09 one first.
0:30:11 Yeah, it’s a really interesting question.
0:30:18 We are just at the tip of the iceberg of seeing AI enter into the energy sector and start to
0:30:20 provide real value in the energy sector.
0:30:25 And I say that and I find it interesting because we were an AI company that launched eight years
0:30:27 ago before AI was cool.
0:30:28 Yeah, I know.
0:30:30 You’re old veterans here, yeah.
0:30:30 We are.
0:30:31 We are.
0:30:34 We’re ancient in the terms of AI in this industry.
0:30:40 But it’s really interesting because we see that we certainly here at Buzz are just starting
0:30:45 to touch the tip of the iceberg of what’s possible from an asset inventorying standpoint to an
0:30:50 optimized maintenance standpoint to a better data-driven decision-making standpoint.
0:30:55 But utilities are just at the beginning of even their data collection journey in a lot
0:30:55 of ways.
0:31:01 So whether it’s inspections, whether it’s demand forecasting, load growth, there’s so
0:31:07 many different ways that AI is helping provide better data-driven decision-making in the
0:31:08 energy space.
0:31:11 And I think there’s just so much room for huge value.
0:31:16 So as it pertains to the next five to 10 years, I really think we’re setting ourselves up for a
0:31:23 lot of exciting ways that innovation can help come in and take over some of these more, I guess,
0:31:29 you know, historical or mundane processes and help the workforce, help the crews make better
0:31:30 data-driven decisions.
0:31:34 So that’s what I’m most excited about, I would say, in the next five to 10 years.
0:31:41 This industry, specifically the utility world, has been very reactive to problems that have
0:31:44 occurred as opposed to being proactive and preventative.
0:31:51 And I’m excited for us to be able to start taking steps forward into a more proactive and
0:31:56 preventative space so that utilities can operate in a, I guess you should say, more data-driven
0:32:00 way as opposed to being so reactionary to everything that’s happening.
0:32:05 And as you look at things like data centers coming online, electrification, you look at renewables,
0:32:11 we have all these massive initiatives that are coming at us at rapid pace, you know, rapid speed.
0:32:15 And utilities, as the, really the backbone of our infrastructure, are trying to figure out
0:32:17 how they can respond to it.
0:32:20 As the backbone of our economy, you know, they’re trying to figure out how they can respond to
0:32:24 all these different initiatives that are happening at such quick rates.
0:32:31 And you look at the age of our utility infrastructure, some of it being over 100 years old, it wasn’t
0:32:37 set up for a lot of these new initiatives like renewables coming onto the grid with bi-directional
0:32:38 passing of energy.
0:32:44 So as you look at the ways that AI can come in and help, it helps be able to leverage mass
0:32:49 amounts of data, turn that into better decisions, and help us as we’re looking to upgrade and
0:32:51 modernize our infrastructure more broadly.
0:32:56 You know, for unfairly throwing you a big answer with a, or a big question, rather, with
0:33:00 a five to 10-year time span, when I heard it say back to me, I was like, we can’t ask her
0:33:00 that.
0:33:01 So much is going to change.
0:33:02 That was a fantastic answer.
0:33:03 So thank you.
0:33:03 All right.
0:33:05 Last question for you.
0:33:06 A little change of pace.
0:33:06 This is an easier one.
0:33:12 Are there AI tools that you are currently really using, enjoying on a regular basis,
0:33:14 whether for work or personal life?
0:33:18 And you don’t have to get into what you’re using them for, but any tools that, you know,
0:33:21 have made your day-to-day a little bit better?
0:33:22 Oh, that’s a great question.
0:33:29 I would say from a, you know, as a company of our size, we are constantly looking, we’re
0:33:30 about 35 people.
0:33:31 Okay.
0:33:31 Yeah.
0:33:35 So as a company of our size, you know, we use a variety of different tools to help us
0:33:38 as we’re looking to optimize and be more efficient.
0:33:41 I would say I’m using AI in several different ways of work.
0:33:47 But one thing that I find incredibly valuable is using tools like AI to help me sift through
0:33:47 my inbox.
0:33:54 I know that’s kind of a funny thing, but if you can get a couple hours back and productivity,
0:33:56 make sure you don’t miss those important emails.
0:34:00 I’m a zero inbox kind of person, but I find it incredibly valuable.
0:34:06 You’ve got a tool hooked up to your inbox now and it just, before you look at it, goes
0:34:06 through.
0:34:09 And does it give you a summary or sort of flag certain messages?
0:34:11 It flags them and helps prioritize.
0:34:12 Yeah.
0:34:14 Prioritize which ones need immediate responses.
0:34:19 And I find, I find not to be incredibly valuable, but you know, as our team here at Buzz, we’re
0:34:25 looking to use AI to help make our process as efficient as possible as we’re an AI company
0:34:27 delivering an AI solution to the market.
0:34:31 Our team is, especially our engineering team is very excited about the ways we can use things
0:34:35 like generative AI and all of that to help improve our work here at Buzz.
0:34:35 Awesome.
0:34:37 Caitlin, this has been a pleasure.
0:34:38 Thank you so much for taking the time.
0:34:39 I learned a ton.
0:34:45 And for listeners who want to find out more about all the work Buzz is doing and the ways
0:34:50 that machine learning and AI are impacting how we get our electricity, places online they
0:34:53 can go, company website, social media, where should they head?
0:34:53 Yes.
0:34:56 You can find us at buzzsolutions.com.
0:34:57 That’s our website.
0:34:58 And then we’re also very active on LinkedIn.
0:35:01 So you can check out the Buzz Solutions LinkedIn page.
0:35:06 We have some great examples of videos of some of our detections, ways in which utilities have
0:35:10 deployed AI, how you can use AI in the utility inspection workflow.
0:35:14 So if you’re interested in learning more, I would definitely check both of those out.
0:35:14 Excellent.
0:35:15 Well, again, thank you.
0:35:21 And best of luck to you and your teams for everything you’re doing, you know, as a person
0:35:23 who relies on electricity every day.
0:35:24 I’m rooting for you.
0:35:25 Thank you.
0:35:26 Thank you so much for having me.
0:35:28 I really enjoyed the conversation today.
0:35:58 Thank you.
0:36:17 Thank you.
Kaitlyn Albertoli, CEO and cofounder of Buzz Solutions, discusses how the company uses vision AI to enhance the reliability of the electric grid by quickly identifying potential issues such as broken components, encroaching vegetation, and wildlife interference from inspection data collected by drones and helicopters. This technology helps prevent outages and wildfires, ensuring the grid remains robust and safe.
Leave a Reply
You must be logged in to post a comment.