This episode of Over the Edge features an interview between Bill Pfeifer and Jeff Necciai, CTO of Duos Technologies. Bill and Jeff dive into how Duos’s railcar inspection portals collect and process data at the edge, then use AI to quickly detect defects. They discuss how this reduces dwell times when trains are sitting for repair and increases the safety of railway workers.
This episode of Over the Edge features an interview between Bill Pfeifer and Jeff Necciai, CTO of Duos Technologies. Bill and Jeff dive into how Duos’s railcar inspection portals collect and process data at the edge, then use AI to quickly detect defects. They discuss how this reduces dwell times when trains are sitting for repair and increases the safety of railway workers. They also cover the other transportation industries that Duos has their eyes on and the challenges they continue to work on in the railway industry.
Key Quotes:
“Our cornerstone product is the Railcar Inspection Portal, where we take 360 degree views of trains at track speed. And we provide the ability to do a remote visual inspection on those, a very detailed remote visual inspection, and we also provide AI around that to automatically detect defects and anomalies on those rail cars.”
“When we're acquiring data from a train that's moving 125 miles an hour, we're capturing over 80 gigabytes of data per second.”
--------
(01:15) Jeff’s journey and how he got to Duos Technologies
(03:39) Overview of Duos Technologies
(05:19) Collecting rail data at the edge
(07:21) What do they use the data for?
(08:46) How were railcars inspected before Duos’s technology?
(12:39) Benefits of remote inspection for the workers
(15:10) Can improving the rail industry reduce carbon footprint?
(17:57) Processing data at the edge railside
(20:23) What data is most valuable?
(22:23) AI and using human-in-the-loop
(25:13) Environmental challenges on the railways
(29:05) Applying this technology to cars and planes
(33:11) The excitement of the railcar engineers
(36:11) The speed challenge
(38:39) New tech that Jeff is excited about
(41:26) Sharing rail safety data with the industry
--------
Over the Edge is brought to you by Dell Technologies to unlock the potential of your infrastructure with edge solutions. From hardware and software to data and operations, across your entire multi-cloud environment, we’re here to help you simplify your edge so you can generate more value. Learn more by visiting DellTechnologies.com/edge for more information or click on the link in the show notes.
--------
Follow Bill on LinkedIn
Connect with Jeff Necciai on LinkedIn
Duos Technologies Delivers Near Real-Time Defect Identification Automation
Narrator: [00:00:00] Hello and welcome to Over the Edge. This episode features an interview between Bill Pfeiffer and Jeff Necciai, CTO of Duos Technologies. Duos has brought innovative edge technologies to a more traditional industry, the railroad industry. Jeff and Bill discuss Duos railway inspection portals. These portals collect data at the edge as trains travel past, then move the data through AI models to quickly find defects and detect safety issues.
But before we get into it, here's a brief word from our sponsors. Over the Edge is brought to you by Dell Technologies to unlock the potential of your infrastructure with Edge solutions. From hardware and software to data and operations across your entire multi cloud environment, we're here to help you simplify your Edge so that you can generate more value.
Learn more by visiting dell. com slash edge for more information or click on the link in the show notes. And now please enjoy this interview between Bill Pfeiffer and Jeff [00:01:00] Natchai, CTO of Duos Technologies.
Bill Pfeifer: Jeff, thanks so much for joining us. I know Duos has been working with Dell, so I've been kind of around your story for a number of years now, and it just keeps getting better.
So I'm really excited to have you on the show and to talk a little bit more about where you came from and what you're doing these days.
Jeff Necciai: Well, first Bill, thanks for the conversation. Really do appreciate it. I'm glad to be here. You know, as far as myself, I really got started in technology, I guess, I guess back in high school, sometime mid eighties, early eighties, you know, probably on a Commodore 64, but my story is not unlike I think most.
Folks that kind of get into this industry and into technology. It just kind of grew from there. By the time I was in college, I was kind of consulting for local townspeople. So it was kind of interesting. Yeah. I mean, that's really the start of my interest in technology. The reason I'm still in technology is because it's both a creative and a technical outlet.
And I think a lot of folks kind of, they don't necessarily recognize that there's a very creative outlet here too. [00:02:00] So it exercises, I guess, both sides of my brain and keeps me busy. It's an interesting
Bill Pfeifer: point. There are a lot of rules to this, but there's a lot of interpretation and creativity. Yeah. So it does allow you sort of a lot of latitude to think right brain, left brain, mix it up and just kind of keep moving.
I like that. So what brought you to Duos?
Jeff Necciai: Uh, so actually, you know, coming to Duos, coming to Duos was kind of fortuitous. It happened quite accidentally in a way. One of my acquaintances used to work at Duos Technologies and we crossed paths in my previous career quite a few times. I lived in Charlotte, North Carolina at the time, and I happened to be in Jacksonville.
You know, I had a business dinner with one of my associates down here. And that friend called me and said, Hey, since you're down there, you might want to stop in Duostech, kind of check out what they're doing. It's pretty cool. And he mentioned they have a, you know, a new CEO and some of the senior management is kind of different than, than at least I was familiar with.
So I stopped in, [00:03:00] I planned on a, I think a, an hour and a half meeting. And I ended up staying for two days. I did. I actually had to go get some clothes for the second day. It was kind of interesting. But I did, when I got back home, I, you know, I explained to my wife, I said, well, I found something we have to move to Jacksonville.
And that was quite a surprise. It was kind of interesting. What drew me here more than anything was the technology I saw. Really up close, what was happening and what the company was doing, especially with the rail car inspection portal. Immediately, I thought, that has so much potential, and I would love to be part of the evolution of that product.
So I sold myself to the CEO.
Bill Pfeifer: Very cool. And you mentioned what Duos does. I know what Duos does because, as I mentioned, I've been kind of on the periphery of what you guys are doing for a while. Could you give our listeners a quick overview of what Duos does and what you do there?
Jeff Necciai: Sure. So, Duos Technologies.
I tell people that what we do is we apply acquisition and [00:04:00] analysis and visual inspection or remote visual inspection to things that move. Now obviously you can appreciate that as a very wide kind of description there, but that's exactly what we do. Our number one cornerstone in there obviously is the artificial intelligence.
But to make that work, we have to go through the stepping stones of all the software that has to be built to make that work, as well as the different technologies that we have to incorporate in acquisition. More importantly, and I guess more familiar to some of your listeners, our cornerstone product is the rail car inspection portal.
Where we take 360 degree views of trains at track speed, and we provide the ability to do a remote visual inspection on those, a very detailed remote visual inspection, and we also provide AI around that to automatically detect defects and anomalies on those rail cars. So, really, that's what Duostech does, and it's applying the same technology, obviously, not only there, but some other industries that we're talking to as well.
Bill Pfeifer: So how [00:05:00] do you actually collect all that information? In my mind, because my mind is a scary place, I just picture like 50 iPhones lying on the track with their flashlights on. And I think it's probably a little more specific than that. So what's, what are the mechanics of how you actually collect it, right?
360 degrees. So you're looking under the train and over the train and all that.
Jeff Necciai: Not trivial. Yeah, no, it's definitely not. There's lots of challenges there. So without giving away secret recipes or anything, I'll tell you, we use an array of areas. And line scan cameras, right? And so folks that aren't familiar with line scan cameras, they basically take one line of pixels from the top to the bottom.
If you take one of those pictures, it doesn't look too impressive, but you put them all together and you got a high resolution image of, of the side or the top or the bottom of the train. And then we have several regular cameras or what I call area scan cameras, like your iPhone. They're not like your iPhone, but they have the same view.
They're angled to give you an oblique view of certain parts of the rail car as well, which are necessary for [00:06:00] certain types of inspection. So that's the easy part. Really what I've said just now is, yeah, we collect a lot of this data with cameras. That shouldn't be surprising, but if you know anything about railroads and where the rail portals are located, light is, you know, it's not readily available, especially when we're under a canopy and under a train, by the way, you mentioned under a train, it's a good example.
There's not a lot of light naturally under a train. We have an engineering department here that not only specializes in the optics and the purpose built optics and the cameras that we use. But we have several patents around the technology we use for the lighting and control of the lighting. We have to get a lot of light on the subject, especially for the train moving at track speed, which could be 74 miles an hour.
Even up to 125 miles an hour for passenger trains, so it's a lot of challenge. And of course we collect other data too. We identify rail cars, we do it through the AEI tag or the RFID. There's a linear speed sensor that we also have patented that detects minute [00:07:00] differences in the train's speed, which are so critical for us kind of synchronizing to our cameras.
So, a lot of collection, a lot of acquisition. And there's a lot of technology behind that. And the engineering here has done a fantastic job in meeting those challenges. HOFFMAN So you're
Bill Pfeifer: collecting these really high, high definition, high speed scans, and using AI, and what are you actually doing with all of that data against the trains?
Jeff Necciai: RICK So we actually did two things with it. You know, those high resolution images, and we'll talk about AI in a moment, but when you think about what the system does, the high resolution images makes it possible for railcar inspectors to actually do a remote visual inspection of those railcars. In most cases, they can see.
Really more than what they can usually see as if they were walking alongside the train. I'll give you an example. One of our modules is called VIEW, which is the Vehicle Undercarriage Examiner. It basically gives you an image of the underside of the [00:08:00] train, the complete railcar. And that's a view that's not normally seen unless you're crawling under a train, and preferably when the train has stopped.
So all of these images become immediately available to the railcar inspector, allowing them to do their jobs. Now, when we overlay artificial intelligence and machine learning on top of that, we can give the inspector basically a second set of eyes, right? Because now we have artificial intelligence that is designed and built specifically to detect certain anomalies and defects that really help the railcar inspector do their jobs and complete their jobs for railcar inspection.
So you're sending
Bill Pfeifer: high definition scans with AI to highlight kind of... And then they can just review the video. What was the process before? What happens if you don't have all these cameras and the AI running? So
Jeff Necciai: if you don't have the rail car inspection portal, that's what you're asking, right? What, what happened a hundred years ago, right?
[00:09:00] And, and honestly, Bill Or even 20 years ago. Yeah, right, exactly. Roughly, it's the same thing. The process of inspecting rail cars and trains, it really hasn't changed that much, right? You generally have two or three mechanical rail car inspectors. They're walking the sides of a train, obviously that's stopped, for the purposes of inspection.
And, you know, they're kind of looking underneath, they're looking at the wheels, they're looking at the bearings, they're looking at these inspection points. They're doing a physical inspection. And just by virtue of the fact the train has to be stopped and, you know, they're out in all kinds of weather and things like that.
The inspection takes longer. It takes a lot longer and there's some safety issues there too. It's really never good to hang around tracks for a long time and trains for a long time. And they got to make sure that, you know, it's protected and things aren't moving. So that's the way it worked. It really hasn't changed in a real long time.
It's just, you're getting human eyes on it alongside the track. So I remember
Bill Pfeifer: from previous interactions that to do a full [00:10:00] inspection was something like six, seven, eight minutes per car. And with the trains that just keep getting longer and longer, we could be talking about six, seven, eight hours or more.
That the train has to pull in, stop, and then people have to walk along and do visual inspection. And that's every trip that it takes, right?
Jeff Necciai: Yeah, yeah, I believe, yeah, on the inbound, yes. So it's one of the things that our technology does. It's one of the values that it provides is you mentioned the train has stopped.
You're absolutely right. It has stopped for a long time. That's one of the dwell times. That's what the industry refers to as dwell time. And it's one of the reasons for dwell time is inspection. And the train has to be stopped if you're doing a manual inspection. If you do a remote inspection, however, the train is still going.
And sometimes if we take a look at a regular standard freight train, it might be going at 74 miles an hour. So it saves a lot, obviously a lot of time. Again, with the artificial intelligence kind of leading the way [00:11:00] on giving alerts to the rail car inspector where to look, where they might want to look because, you know, AI found a specific defect here on this car or somewhere else on another car, they can actually go in and see that.
It just makes the whole process a lot faster and a lot safer. So we
Bill Pfeifer: need that from a safety perspective. But then I have to imagine everything seems to be getting faster, right? Like Amazon and other companies are giving me one and two day delivery. Now they're cutting it down to four hour and overnight delivery.
And anything I ship by train is now going to sit there for 10 hours before it even starts the journey, which seems kind of excessive, right? So cutting out that 10 hours probably makes it a lot easier for companies. To adapt to still using trains for shipping because it's going to be so much
Jeff Necciai: faster.
Absolutely. I mean, if you look back over the recent history of the railroads, positive train control and all of the other technical advances they put in place, there's a lot of reasons. They're very competitive in the railroad industry. All of those [00:12:00] folks, all of those companies are, are looked to, to provide punctuality.
And capacity. Those are the two things that railroads talk about. How can we get more there faster? And quite honestly, cutting out a little bit of that time that is used in inspection, you know, really helps in that. It also makes for a safer journey as well, right? With the additional technology that's in place.
Remember we said second set of eyes, for example. With the additional artificial intelligence that's in place. And very often things are identified that, you know, it just wouldn't normally be identified. Again, safety and efficiency. That's the mantra of what we're talking about as far as the solution is concerned.
Right. And
Bill Pfeifer: hot summer, cold winter. I'm sure people aren't very excited about spending extra time to make. Double sure that that train's safe and crawling around underneath it and things like that. So hey, I can do that for them. And that's, that's
Jeff Necciai: much nicer. Absolutely. I'm sure it's much better to perform that inspection in an office environment rather than trackside in Montreal in [00:13:00] the dead of winter.
It probably makes it a
Bill Pfeifer: lot easier to hire people too. Here's what the job's going to look like. So I don't know if you've looked, you've probably paid more attention to the human aspect of this. I'm just thinking early in my career, I worked in a government facility and they were looking at in 10 years, something like 80 percent of their workforce was set to retire.
Because it was just by and large an old demographic and young people weren't flocking to, Hey, I want to work at
Jeff Necciai: a government facility. Yay.
Bill Pfeifer: And so, you know, I think back and probably everybody I worked with there is already retired and just out of the workforce. I can't imagine that the railroad industry, especially the inspection industry is any different, right?
It's probably not a huge draw for people. I never said when I was You know, in third grade, I want to be a rail car inspector. I don't know where people come up with that as a career path, but I would imagine that it's probably pretty hard to find
Jeff Necciai: them. Yeah, [00:14:00] for the railroads, I know that they're always interested in attracting, you know, younger folks.
I think a lot of business want to attract younger folks to their industries. And I don't think the railroads are any different. I do know with all of the technology that the railroads are incorporating, not only ours, but technology across the board, I know it's becoming more attractive to younger workers.
So I'm sure that's one of the reasons. It's not the reason for adopting the technology, but it's certainly one of the benefits from having newer technology. Well, for example, we're talking remote visual inspection, the workforce that's coming to work now, they're Kind of semi familiar with computers. They understand how they work.
They've used them before. There's other things going on. And it's just more attractive to a younger workforce when they have all that technology. Much more like
Bill Pfeifer: playing a really boring video game than slogging it out through. Six inches of snow on a
Jeff Necciai: frozen drain. Right now, you know, let's be aware, Bill, you know, somebody still has to fix them.
Right. So when an anomaly, a defect is [00:15:00] found, we're always going to have mechanics that go in and actually replace a part, replace a component, or go ahead and fix them. As a matter of fact, for a long time, or at least for a time, fixers. And that's really the truth. You know, we're able to find more defects.
Which means more repairs.
Bill Pfeifer: And probably a lot more efficient work for the few people, the relatively few people who are up to speed on how trains really work, right? You don't have to walk around trying to spot errors. You can actually have somebody tell you where the error is and then you fix it. That would be much more rewarding, I think, as work.
Yeah,
Jeff Necciai: absolutely.
Bill Pfeifer: So. Have you looked into what that does in terms of carbon footprint, right? Like hauling freight across the country. You can do it over land trucks and things like that. You can do it by planes. Very different from a sustainability perspective. I know trains are really, really efficient, steel, steel wheels rolling on steel [00:16:00] rails, right?
If it's not. If you're helping speed it up, which is fantastic, or too dangerous, and you're making it safer, which is fantastic, then that's a pretty great sustainability story, whether it's something that's really a selling point or not, I don't know, but it seems like that would be kind of a nifty add on.
Jeff Necciai: Yeah, it is, and it, you know, it makes total sense when you think about it, too, because we look at mass transit, we don't give it a second thought. We know that that's more efficient, pulling a busload of people across town. Makes more sense than having all those cars on the road. It's more fuel efficient and so forth.
And obviously it's friendlier for the environment. The same is true, maybe even to a greater degree. I know recently I read a study that says that hauling freight. On a train, it reduces greenhouse gas emissions by 75 percent as opposed to hauling that same freight via truck on the highway. And I can believe that.
I mean, I can believe that because look at what we're doing. It is, you're taking advantage of the mass economies of scale, so to speak, right? We're moving all that freight with, you know, [00:17:00] one or two or four engines, maybe. But it's still, you're benefiting there, I think, in a large way. And let's not forget that, yes, you have the train.
Yes. It's on mostly a flat surfaces on rails. There's very little friction. So. You have some energy efficiency already built into the system, but they've been working on alternative fuel sources. They've been adopting all kinds of things, really interesting and bleeding edge kinds of technologies that, you know, really have a good effect on what they're doing as far as fuel efficiency and the.
Carbon footprint is concerned. It's
Bill Pfeifer: not an industry that we tend to think of as super high tech, but then you can make huge, huge advances with relatively small investments in tech. And I say relatively, because I'm sure you don't think that it's, you know, building all of those, all of those things out in the middle of no place is probably not quite
Jeff Necciai: minimal.
Right, exactly. Yeah.
Bill Pfeifer: So it's, it's using edge computing, which is of course all about the data. You had talked about the types of [00:18:00] video data that you collect, but do you actually process it right there on site at the edge? Do you backhaul it? Do you have a split thing of, you know, you backhaul some of it, or how do you actually handle that data?
Jeff Necciai: Yeah, we, we absolutely do all of our acquisition and our analysis or our processing is all done at the edge. It's done in what we call EDC, Edge Data Center, and the Edge Data Center is located trackside. So the reason it's done all there is because what we're looking at here is a mission critical system.
Our customers require and demand timely results from the system. So in practical terms, think about it. If there was a critical defect. On a rail car, and our system didn't tell you about it for two hours, that train's off going somewhere else by now. It might be a little too late for that, you know, you have to pull it off at the next stop.
We've worked very hard to get this information out to the rail car inspectors and the railroads as quickly as possible. And what does that mean? [00:19:00] On the average, our detections, and that means all the way from the point of image acquisition through analysis. All the way to the point where we make it visible to the customer and alert the customer of a defect.
We've gotten it down to about 1. 3 minutes on the average. That's not really bad at all. And actually, that was quite an improvement over the last six months. I think we were hovering around five to seven minutes on the average, which, by the way, is still great for the industry as a whole. But by doing edge processing, it allows us to do that.
Bill Pfeifer: And especially relative to what there was in the industry. Prior to that, right? Even five to seven minutes is like scorching real time, crazy fast. So of
Jeff Necciai: course, you know, I mean, look a lot of, I don't even want to say competitors to us, but a lot of similar industries that utilize AI for similar things they do for economy purposes.
They put the analysis in the cloud, so all of the acquired images and everything, they're kind of backhauled to the cloud, [00:20:00] and that's where the acquisition is done, and, but that takes a lot of time, especially with the size of data that we have, the size of the images are, are huge, it would add several minutes to those detections, so it costs a little more, it takes a little more work, but we prefer to do that processing at the edge, and then alert our customers as we need to.
That makes sense.
Bill Pfeifer: Yep. The track sits empty for a long time and then a train comes by and you have this massive burst of data. So rather than backhauling that someplace slowly over time, processing it all at once to get the speed of response would be A lot more interesting and a lot more useful to the customer, I would imagine.
So, what do you consider to be your most valuable data, if that's not trade secrety kind of stuff? And how do you decide what's valuable data that should be kept, what's ephemeral data that you just process and dump? How do you handle that data
Jeff Necciai: management? Sure. We look at that quite often, but we don't separate it in terms of categories of data.
It's really the timing of the data. If you look at what our system does, [00:21:00] it acquires, let's say, mostly high resolution images, but, but some metadata along with that, and then it subjects it to AI and it makes it visible to rail car inspectors. Once that job is done, once the AI has, and by the way, yes, that's, that's the point where all of the data we're collecting is critically important because it hasn't been seen yet.
But once the AI has analyzed the data and once the rail car inspector has basically inspected the rail car or inspected the train, then the importance of that data, it slips down by a notch or two. And we do have segregated ways that we save it or make it available. It's important for a period of time, and then it's important, it's kind of relegated to things like looking at historical patterns and possibly making it available for predictive analytics and so forth.
And
Bill Pfeifer: probably retraining the AI models and things like that. You've got to have a bunch of data for that, I would imagine. Yeah,
Jeff Necciai: we do. So, what we do with our models is it's a [00:22:00] continuous learning pipeline, and we actually have a It's almost a third category of data, right? Because there's a little bit of a different path that we take with that data.
We do quite a few things to ensure that our models are always kept up to date and always retrained when they're needed to. We do utilize human in the loop for new models. So, for example, when our AI department, our AI engineers produce a new model for a specific defect for a while, and it could be a period of two months or 90 days or something like that, it is subject to human in the loop.
Where each and every detection or every non detection from that model actually has human eyes on it from our 24 7 data center. They look at that and they actually make a human decision saying that, yep, the AI was correct or no, the AI got this one wrong. And so those results are gathered continuously, kind of under a separate stream of data, and it's fed back into our continuous learning system so that those [00:23:00] models can be retrained.
We do that, like I said, for a period of time, maybe 60 days, 90 days. And then we obviously continuously monitor it, monitor it after that for performance, but we think that that really, really helps. It is a supervised learning model, but we think that we get the most accurate results that way early on putting human eyes on it.
Bill Pfeifer: And when you're talking about safety and system failures and component failures, you definitely want that extra bit of safety, just to make sure that we're not trusting the AI too much before you show for sure that you can trust it. That makes a lot of sense.
Jeff Necciai: Right. Well, the other thing it does too, is it prevents our customers, our railroads, our rail car inspectors.
From getting too many false positives, right? Because by and large, I think that's what a lot of the new models do when they're first released. You put them out in the environment that they're kind of running wild now. And what you realize is you have more false positives and that's a problem for the railroad because they may take action on something that's really not [00:24:00] an issue.
It's an
Bill Pfeifer: interesting thing. There are lots of conversations going on in various parts of the industry about IT versus OT. And, you know, IT is all about move fast and break stuff and modernize the business and OT is all about, but you have to run the business and it has to be up and it has to be safe. And finding the balancing point between that causes immense friction between the two groups.
And it sounds like you're doing a really good job of finding that middle ground and keeping that line. Pretty clear for your customers. So they don't have to worry about, you've got the ITO angle and the OT angle both covered, so they don't have to think about it. That's awesome.
Jeff Necciai: Yeah. And once again, it's part of the nature of the solution for that as well.
So we see nothing but, you know, more opportunity for that in the future as well. Right. So I'm still
Bill Pfeifer: thinking about these sheds with the cameras, which are totally not 50 iPhones lying on the track with their flashlights on. It's just a funny mental picture. I can't get that out of my head. So. You have all this technology that's out there, [00:25:00] kind of in the middle of nowhere, and I would imagine that...
You've got some challenges with power and with environmental. It gets really hot wherever, and in the summer, and it gets really cold in the winter. You'll have some component failures. I assume you probably don't have these things manned because who would want to sit there just staring at technology?
Not being broken. So how do you manage all of these sites that are far out there, right? When, when you have failures, how do you deal with that? When you have new technology that needs to go in or a camera gets knocked off its perch and isn't looking at the right thing now, how do you handle the mechanics of just being distributed like that?
Jeff Necciai: So, yeah, let me kind of address that in a couple of different ways. First, from a micro standpoint, just looking at one rail car inspection portal, the subsystems that we have. In place, you know, we mentioned the view, the vehicle undercarriage examiner before we have the overview. That's the oblique vehicle under carriage examiner and all the cameras and the lighting and the [00:26:00] sensors that we're talking about.
They've all been ruggedized over time. It doesn't happen all at once. It's engineering. It's re engineering, it's designing, it's going back to design, creating version one, two, three, four, right? And this is just normal course of action. It's the normal course of evolution for that type of equipment. But the extremes are the extremes.
We have sites in Mexico, we have sites in Canada, we kind of run the gamut on temperatures. And even without the weather and without the temperature, you know, these are trains and they make a lot of vibration. There's things like track pumping and things like that that actually kind of move certain small parts of the earth.
So sure, there's challenges. Again, our engineering department has worked tirelessly to make sure that these components are regarded as mission critical. And they do stand up to those environmental things. Now, that being said, these are cameras. They have lenses. Think about the vehicle under carriage examiner sitting under, under a train or at least on an empty track.
And yes, we have a [00:27:00] canopy, but when a train comes through, you know, it can pull rain, it can pull sleet and snow in with it. It's kind of like a vortex of weather that comes in that that train pulls in. So, you know, a lot of that kind of lands on the vehicle under carriage examiner. We've built in protections for that, not protection for landing on it, but what happens to it after that.
We've outfitted them with air knives to blow the debris off of them and things like that. Even with all of those, the protections like the heater and the heater blower, and we have for some of that equipment, even with all those protections, every once in a while, it does require some human care, right? So, depends on the weather patterns, but every once in a while, somebody usually from the facility, and they do get visited from time to time, if not by us, then by the railroad clients themselves.
They simply go off and they wipe off the lenses. They're not the actual lenses, but the glass covering the lens. We're good. So the other direction I wanted to explain is, yeah, these are in [00:28:00] remote locations. All of them, really. They're really in remote locations. But we have a 24 7 service desk. We monitor all of those locations 24 7.
It's part of the service that we offer. We're looking for everything. We monitor the condition of every single camera and everything, every single sensor. And we make sure that those systems are up and running at all times. And we're very proactive if anything should be wrong. Because again, these are mission critical.
We don't want to miss a train. We don't want to miss a rail car. We don't want to miss a detection. So that, that's why we've committed to 24 7 service on that. Very cool.
Bill Pfeifer: Very cool. I love the solution. So early in this, you said that you were in data acquisition of things that move. Yeah. And then we started talking about railcars, but that sounds like maybe there's other stuff that you're working on or working toward.
Again, not toward trade secrets or, you know, secret future stuff, but are there other places that you're looking to apply this [00:29:00] technology? I mean, it's a steep learning curve and you've gone through the steep learning curve, so you might as well, you know, expand that to everything that you can reach, but
Jeff Necciai: are there plans for that?
Yeah, absolutely, Bill. And you're right about the steep learning curve. We have gone through quite a steep learning curve. We've paid our dues, so to speak, right? And we've done some really cool things along the way, but it does relate to other industries. And we know that we're not blind to that. One of the other industries, for example, and I don't know if we can classify it as an industry per se, but our plan is to take the same technology.
We're releasing a truck inspection portal at the end of the year, beginning of the next year, releasing a truck inspection portal. Which does for over the road trucks, really the same thing that the railcar inspection portal does for the rail trains. So there, you can make a correlation to what we do.
What are we looking for on trucks? Well, you know, we're looking for certain damages. We're looking for holes in ISO containers and 53 foot trailers. We're looking to make sure that certain things are... The way they should be, like the locking rods on the back [00:30:00] of the 53 foot trailers, and we're looking at the landing gear on chassis, those are all inspection points, if you will, that the truck inspection portal will kind of identify.
It's very similar to the railcar inspection portal in many, many ways. But we do feel that it expands our scope a little bit on what we're doing with the technology. And certainly we're not stopping there. We've had lots of, well, I shouldn't say lots. We've had several conversations with some R& D companies.
That are focused on airframe inspection as well. So, hopefully by this time, maybe halfway through next year, we'll be talking to you about what we're doing with airframe inspection because that would be kind of cool too. Yeah,
Bill Pfeifer: I could easily imagine on the way to takeoff or on the way to landing. Yeah.
Just kind of roll the plane through one of these giant portals. If you could throw a de icer in there, that'd really help save everyone time in the winter. So as AI continues to [00:31:00] develop, where do you see that going just overall? And how do you see that changing transportation at the edge?
Jeff Necciai: I know in our business and other similar businesses, our catalog of AI detections, I think has around 43, 44, we call them detections, but there are 43 or 44 models that specifically look for inspection points that are regulated by regulatory agencies.
We're expanding that model. I think by the end of the year, we're looking for 56 to 60 models by the end of the year. So our company is going to continue to produce those models all focused on, again, regulatory compliance. So they're very important kinds of detections that we're looking for. In addition, we have added AI at other parts of our business and in the railcar inspection portal that don't actually look for defects on railcars.
Instead, they help our company [00:32:00] look for, well, let's say defects within the railcar inspection portal. There's garbage on the lens or maybe we're not getting the right lighting on a certain view. That's all monitored and continue and growing actually, but it's being monitored by AI and machine learning.
So we're excited about that. And we see that that expanding into some other parts of our business as well. And I think, look, we've all heard recently, you know, there's chat, GPT, and all of the open AI things that are happening right now. I think we're going to find more and more ways to make use of AI in much better ways.
So I have to ask,
Bill Pfeifer: do you have a favorite part of the inspection process, like some particular part that was a really difficult problem to solve and you're really proud of it? Or something that you think, wow, I can't believe we can already do that. That's amazing. Just like sci fi kind of stuff, because there's so many of these things that you get one little nuance and you're like, wow, that just makes all the difference.
It's amazing. Is [00:33:00] there like a particular part that stands out for
Jeff Necciai: you? You know, bill, actually, there's, there's so many. I'll tell you, I wish I had one of my, uh, well, this is one of the things we didn't mention. I, I wish I had one of my mechanical rail car inspectors here. And when I say mine, they're, they're actually mine.
I like to tell customers, you know, we're a technology company. We don't own a rail car. We don't own a train, we don't own a single inch of track, but we have mechanical rail car inspectors on our staff. And these are mechanical railcar inspectors who've worked for the railroads with, you know, 20 some years of experience each.
But if you ask them, they're constantly amazed by what our data scientists can do. Now, they work very closely with the data scientists. They explain exactly what we're looking for and why we're looking for it and why it's a safety issue or why it's going to cause issues down the line. And so they work together as a group.
The railcar mechanics, they are our subject matter experts, and they work to produce these models really together. It's a collaborative effort. But when the model is actually [00:34:00] identifying these things, and it could be something as simple as a bent brake beam, or, you know, something having to do with the roller bearings, or flange, which is extremely important.
I've heard a lot about the broken flange. To see this amazement on their faces is amazing. Now, for me, it's... Yeah, I like it all, Bill. I mean, I think it's all fantastic, but I don't have the benefit of being a rail car inspector and I don't have the benefit of knowing how it was done before. And the rail car inspectors do, so to really see them light up makes me feel good.
And I think it's pretty cool, but no, I actually, to be honest, not one particular one stands out. I could tell you a story about, we have human detection, right? So we have one of our algorithms detects humans riding a train. These are freight trains. Those, these aren't passenger trains. So it's not a very advised thing for, for, it's a, you know, a very bad safety issue for humans to ride on the outside of a 74 mile an hour train.
And a couple of times [00:35:00] early on, our algorithm picked up graffiti on the side of a couple of these cars. It's very popular to put graffiti on cars, the rail cars, and some of it is extremely good. I mean, the graffiti is extremely good. So there was one time where. Our algorithm picked up a rendering of Al Pacino, you know, it detected, it said, Hey, we have a human there, you know, but you look at this, it looks like a picture of Al Pacino and of course it was just graffiti, but yeah.
So that's an interesting little thing. It's surprising when that happens. I love it.
Bill Pfeifer: Using AI to identify budding young
Jeff Necciai: artists. I never thought of that. See, you've opened another market for us, Bill. I appreciate it. There we go. Railcar art.
Bill Pfeifer: I love it. So you're helping remove a lot of concerns for your customers in terms of, you know, what's going on with their trains and keeping them on the line [00:36:00] and reducing those dwell times.
What do you get concerned about?
Jeff Necciai: Well, one of the challenges, we've talked about the engineering challenges and we talked about optics and lighting and the environmental conditions and how we've met those challenges. We continually meet the speed challenge, right? So we went from acquiring data for freight trains at let's say an average of 60 to 70 some miles an hour.
And that's a lot of data that we're acquiring all at one time. It's from several different perspectives. It's high resolution images and there is no queuing mechanism. We can't really queue it up. So we have to acquire it and store it. So recently, not recently, past two years, we've been looking at and working with passenger trains that go up to 125 miles an hour.
So that exacerbates the speed and the acquisition problem. So one of the things that we're doing to kind of help us through that. Ingestion, that data ingestion challenge [00:37:00] is we work with some partners. Well, I mean, Dell, for example, we work very closely with Dell to kind of help us build those solutions that will work for us and give you an example of what one of those solutions is able to do for us.
I mentioned, mentioned passenger trains when we're acquiring data from a train that's moving 125 miles an hour, we're capturing over 80 gigabytes of data per second. One of my engineers puts that in perspective, he said that's enough to fill up the average computer, the average home computer in four seconds.
So 80 gigabytes per second is what we need to acquire. We can't lose any of it. And that's what we're acquiring and analyzing. So if anything, I don't know that it's a concern. It doesn't keep me awake at night yet because we've met those challenges so far, but it has required some effort. It has required some redesigning of existing systems.
And like I said, we have benefited from our partnerships greatly in that respect, but I see that requirement, I see that [00:38:00] growing, that's not going to go the other way. Trains are going to go faster. We set our sights on European markets where most of the trains are faster. So we're going to have those challenges in the future.
So we're just going to keep pressing the pedal as much as possible. So
Bill Pfeifer: that was concerns. We can't end on a, on a down note. We have to go, you know, kind of upbeat from there. As we move forward. What parts of, of this solution, of the evolution of technology, of edge computing are you excited about? What are you looking forward to next as the future for Duos Tech or for you and your
Jeff Necciai: career arc?
Well, we have, okay. So first of all, for Duos and our product line, we have quite a few changes that are coming up over the next, I would say six months, six to eight months. We're changing lots of aspects of the way the rail car inspection portal actually works. The internals of it and some of the optics and some of the lighting.
I can't get into much specifics right now, but obviously we're, we're reaching for continuous improvement all the time, and there's gonna [00:39:00] be a flood of that over the next six to eight months, so we're excited about that. On the AI front, some of the technologies that we're using now, again from new advancements from NVIDIA that we're taking advantage of being able to load balance GPUs.
GPUs are very expensive, obviously, and sometimes you have limited chassis space for GPUs, so. You know, how to actually load balance that correctly because our number of models are continually increasing. We're looking at a time where we don't really necessarily have to play catch up. It just becomes muscle memory to go ahead and add another module.
So, I'm excited about those changes because it just makes our job easier when we utilize the new technologies.
Bill Pfeifer: Excellent. I love this idea of... applying the newest technology to one of the oldest industries in a way that's not invasive, right? Like you're not changing the trains. You're not making them less reliable.
You're just adding something on the outside that makes the whole thing flow faster and [00:40:00] be safer. That's just, that's a really great business model. And it seems, it seems like a clear win. I love that Duostech is doing it. I love the story that y'all have. And then, you know, expanding that to other forms of transportation completely makes sense.
It would be interesting when you start to get down to automobile safety, right? You know, identifying cars that shouldn't be on the highway moving at the speed that they're at or something like that. That becomes a lot more contentious and, you know, mass produced. And there's, there's a whole lot of compute that we'd be talking about there.
But interesting business model. I love it. And I, I love that story of how Duos has evolved into doing this. So I really appreciate your time today. Is there anything else that you'd like to add to the conversation? Anything new that you have coming in the relatively near term?
Jeff Necciai: Well, first of all, thank you for the comments.
I fully agree. There's lots of new features coming, obviously, or, you know, it's constantly an [00:41:00] evolution of software and hardware and system. We're all excited about that. That's why we come in every day. But one of the things that really have us excited now is because we're going to begin sharing basically rail safety data with the industry as a whole, right?
So when you take a look at a rail car. It's on a railroad, which is good, but that rail car is owned by an owner and it's operated by an operator that may be separate. And right now, because of the limitations of technology, right now they don't have visibility to the same information that the railroad does, which would be very beneficial.
Again, it's safety information. So through a subscription offering that we have launched, we're offering that to rail car owners so they will be able to see. Defects and anomalies that our AI has picked out, they will be able to see those high resolution images as well, which will allow them to maintain their fleet, make it safer and more operationally efficient.
That's one of the aspects that we're extremely excited about, that has kind of become the [00:42:00] talk around the hallways here, so. That's
Bill Pfeifer: very cool. So really tying all of the data that you're collecting and already sharing with the maintenance side of folks, but then sharing it back to the actual rail car owners so that they can track what's happening with particular pieces of their
Jeff Necciai: fleet and their fleet as a whole.
And if you can imagine, I mean, if I'm a rail car owner, I'm given the opportunity to keep that rail car safer, to kind of keep it more well maintained. That's better for the railroads because they don't have delays from a rail car that's not working correctly, so. Right, and then
Bill Pfeifer: they can potentially keep their fleet in service longer because they can watch trends and, you know, do you need to overhaul a part of it before it just catastrophically fails or something like that.
That's fantastic. I love it.
Jeff Necciai: Win
Bill Pfeifer: win win. That's a really great addition for the technology. Totally makes sense. I love it. That's just so good. Thank you. Continuing on that great Duostech story. So Jeff, how can people [00:43:00] find out more about you, find you online, keep track of what you're up to, what Duostech is up to, and follow the story as it continues to
Jeff Necciai: develop?
Sure. I mean, obviously you need to go to the website, www. duostech. com. And you can find us on obviously LinkedIn. You know, if you're on the website, there's a phone number. Give us a call. We love to talk about ourselves. And if you need, we can put you in touch with not only salespeople, but technicians. We do it all here in Jacksonville, Florida.
So we're all under one roof here. We're one big family. So. We can talk to you if you have some questions. I love
Bill Pfeifer: it. Jeff, thank you so much for your time today, and for
sharing your story with us.
Jeff Necciai: Thank you. I appreciate it.
Narrator: That does it for this episode of Over the Edge. If you're enjoying the show, please leave a rating and a review, and tell a friend.
Over the Edge is made possible through the generous sponsorship of our partners at Dell Technologies. Simplify your edge so you can generate more value. Learn more by visiting dell. [00:44:00] com slash edge.