Winn Hardin: [00:00:07] Hello everyone and welcome to Manufacturing Matters, where we talk about the manufacturing technologies and trends that are reshaping the manufacturing global marketplace. My name is Winn Hardin. I’m cohost of Manufacturing Matters and managing director of Tech B2B Marketing. And as usual, I’m here with my good friend Jimmy Carroll, vice president of operations.

Jimmy Carroll: [00:00:25] Hey everybody.

Winn Hardin: [00:00:26] And today we’re lucky enough to be with Mark Willingham, Headwall Photonics CEO, and Christian Felsheim, Headwall Photonics general manager of machine vision. So Mark, Christian, thank you so much for joining us today.

Mark Willingham: [00:00:39] Thanks for having us. Very excited to be here.

Winn Hardin: [00:00:41] Wonderful. So now I’ve known Headwall Photonics for 20-plus years, back from when I was managing editor of Photonics Spectra, so long ago when I had hair at that point. And I’ve been covering your industrial machine vision and deployments for those decades. And we’re familiar with material differentiation using hyperspectral and multispectral imaging systems with food grading applications, with a whole number of advanced imaging applications used in industrial environments. But before we go and start having specific discussions about the technology and a lot of the interesting changes that are happening at Headwall Photonics, I was wondering if Mark or Christian, you guys could just chime in and give us a quick overview about what Headwall does, where you’ve come from, and the capabilities that you’ve been bringing to the industrial automation market for quite some time.

Mark Willingham: [00:01:35] Yeah, sure. I’ll kick off. Again, thanks so much for having us here. It’s really exciting. This is a fun podcast and we’re glad to be part of it. And at Headwall, we’re doing something very simple, which is building the preeminent supply, becoming the preeminent supplier of hyperspectral technology and optical components in the world. It sounds so simple, but the time has come, and we are the ones. So we break into three main business units. We have our own optical components business, where we develop kind of optical gradings and micro-lens arrays and some of the optical components that serve a wonderful market and growing market and add to our portfolio that way but are also many of the critical components used in hyperspectral imaging systems. So we are the masters of our own destiny when it comes to the optical components that are so critical for perfection in the markets we serve. That bleeds up to the hyperspectral systems. And as you say, it’s been decades that Headwall has been in that market. But it’s very different company today than it was even five years ago. Now we’re split into two sections. One is the better-known remote sensing, where whether it’s satellite or fixed wing or drone, we’re bringing hyperspectral data to users looking at agriculture or mining or many other research-type topics. But really the exciting part is that, and what Christian is here to help talk about is the fact that hyperspectral is now in what we call the machine vision market, where we’re using a hyperspectral imager in the same way you might have used a camera in some kind of volume production. That could be things whizzing by on a production belt or inside automated equipment that’s performing soldering or other electronic componentry.  We do both of those. We do them at scale and we do them at speed.

Christian Felsheim: [00:03:34] Yeah, and we are working closely with a lot of applied research partners to develop further applications. We are in quite a bit of industrial applications already, but we really feel that the time is right to scale it and to drive it out in many more applications, whether these are in food safety and quality, pharmaceutical, agriculture, but also infrastructure applications, inspection applications, or environmental monitoring, recycling, whatever, you name it. Just general analytical process control. And across our application teams, which we are now having in the US and the Netherlands as well as in Germany, we did just over the past few years more than 600 application studies to drive this technology out into the field.

Winn Hardin: [00:04:26] So when we say application studies, we’re talking about a lot of real-world deployments mixed into that group, correct?

Christian Felsheim: [00:04:32] Well, not all of them certainly makes it in real deployments. We are working with partners to utilize this technology, and that certainly gives us a wide understanding on how to use it and where to use it best.

Winn Hardin: [00:04:44] Right. I know I’ve seen some recent stories about, I believe, real-time water monitoring for pollutants and other contaminants.

Christian Felsheim: [00:04:52] That’s a very interesting one indeed, yeah, just coming up.

Winn Hardin: [00:04:56] Yeah. And then of course the old-school-type application recently about food grading. I think it was related to oranges. I think I saw something recently.

Mark Willingham: [00:05:05] It’s interesting. We think old school like that’s been happening. But a lot of that’s been happening in research or looking for things that you might see. And I think when you fast forward today, what you see is that these are starting to deploy or are deployed at the scale the industry needs to be useful across the value chain. Things like fruits have in their value chain from grower to consumer, there’s many points along the way where information is critical and they today might destructively test or have destructively tested to understand the how that produce is progressing. And now hyperspectral is in use in ways that are helping with incoming inspection or outgoing, helping with the status of product that’s moving through the value chain. So it’s out there, and in many cases in the fruit segments replacing destructive testing that was slowly bleeding away the very product that we’re trying to get to the end market.

Winn Hardin: [00:06:07] I look forward to hearing more about agtech. But before we do, Jimmy, you had a thought on just the general technology, setting it up for us.

Jimmy Carroll: [00:06:14] Yeah, well, I like asking folks like you guys about where hyperspectral imaging is in the industrial space. And what I mean is, not that long ago, hyperspectral was considered something that was really only for the lab. It was large. It was prohibitively expensive, whereas the use of technologies like RGB and machine vision are very mature and there’s a lot of information, but there’s a lot of information that can’t be picked up using RGB. And so you touched on this a little bit. But what are some common manufacturing challenges that are suitable today for hyperspectral, where conventional, visible RGB can’t be used, and can AI play a part in that?

Christian Felsheim: [00:07:01] Well you are touching a wide range of of topics. Let me let me address that with the RGB cameras. I mean, certainly when we look at the machine vision market, actually most of that is pure black and white because it’s just identifying: Is there an object? And if there is an object, where is it? That’s fine with black and white, but just look on how much you’ll get out of an RGB image compared to a black-and-white image. Just think about an apple. When you look at an RGB image, you can tell whether it’s ripe, whether it already has red cheeks, or whether it’s maybe even too ripe because it starts developing brownish spots. That’s what you get out of an RGB image compared to a black-and-white image. Now compare that with a hyperspectral image. The amount of information increases. You’re no longer having just three layers of information but 200 or 300 layers of information. That gives us much more value to provide to the applications, like, let’s just stick with the apple. We can tell about the sweetness, the Brix value. We can tell about the acidity of the apple, and we can even identify how firm it is. Is it good to go into storage or would it need to be sold right away? So that helps to really generate value for our customers and also reduce the amount of waste during the production. And that’s what really drives now the way into the market and the value it can create for our customers.

Winn Hardin: [00:08:39] So we’re going far beyond just using near infrared or shortwave to be able to detect bruises below the skin. And we’re actually getting into grading. which gets right back to increased yields across the board. Better pricing structures for your customers to be able to know where they could charge prime versus lower-grade fruit, meat, whatever the case might be. This makes me think about when you mentioned just now Christian about the 200 to 300 wavelengths. So when I was first getting involved with HSI or multi-wavelength, the traditional application development seemed to be, usually because of the cost of the hyperspectral imaging system, you get an HPCI system, and if you’re going to do volume deployment out there, you find your signature wavelengths, you optimize it into a multispectral system, and you build the systems out. Is that still the way to market today or how has it changed?

Christian Felsheim: [00:09:35] Well, in several ways. Actually, we are currently at a sweet spot of the technology, where we clearly benefit from the recent developments, bringing down price and raising the performance. We are seeing that on the sensor side, we are seeing that on the algorithmic side, with all these AI algorithms coming up and getting packers on board. We can certainly have some more discussions about that later. We really benefit out of that but also out of this processing machinery, which is now becoming really fast and make that available to drive into the market. So we really benefit out of that, and that helps getting this technology to market.

Jimmy Carroll: [00:10:15] I hope for your sake that if you have friends and family that ask what you do for a living, you tell them that if they happen to bite into an apple and it tastes really good and it’s not bruised, they have you to thank.

Christian Felsheim: [00:10:28] And more and more of such examples coming up, indeed.

Mark Willingham: [00:10:33] Christian really hit the head. We’re at an inflection point. And I think that’s where Christian was driving. Where edge computing allows us to crunch those numbers faster, to get the data to use faster. We have AI that helps us with our own software by the way, with the acquisition of perClass, we’ll talk about, embedded fully in our solutions, where we’re able to take the data, do the computing, and interpret in a way that’s immediately useful. And this is why if you go back just five years, those didn’t exist in a way that was available, meaningful, and embedded now like we are in our system. So it’s really a new world in terms of the ability to deploy into more commercial areas. One of the things that I didn’t mention in introducing us as the preeminent supplier is the fact that we’re also in partnership with Arsenal Capital Partners. And that gives us a huge opportunity to grow our businesses that exist today but also grow inorganically and add the businesses we need to become the integrated supplier of the future. And we’re really unique in that space. And maybe I’ll just continue and chat a little bit about that. I mentioned that we make optical components. Well, in addition to the components that we made, Winn, to your point of years ago, we bought a company called Holographix.

Mark Willingham: [00:12:05] They make replicated gradings and micro-lens arrays and other very high-end optical components. That’s a much higher volume and more repeatable grading that we now use in our systems, in addition to running that business. So we get that experience and that manufacturing base. Overall we’ve moved from two locations in Massachusetts to five locations in three countries in terms of acquiring Holographix and then the software company perClass, and more recently inno-spec in Nuremberg, Germany. So now we have a base of supply in both hyperspectral, industrial optical components here in the US and a software team in the Netherlands. And what that additionally gives us to do beyond the hardware is we can now attract and add talent in any of those areas serving any of our business units. So it really engages us in a way that really nobody else is right now, where we can rapidly move on the technology from component to imager. In the software, we’re developing it constantly within our own confines and in the manufacturing or software or sales or any talent, we can house them across a huge geography closer to our customers.

Winn Hardin: [00:13:32] If I could follow up with a question on that real quick. So you mentioned the micro-lens arrays, of course, and this helps, if I’m understanding you correctly — and please correct me if I’m mistaken – that hyperspectral imaging produces a lot of data in a traditional sense. The data cube itself is intense. You try to use hardware elements as much as possible to reduce and expedite that throughput, right? And then of course enhance software through artificial intelligence, other improved pattern matching to be able to turn the measurements into insights and intelligence that can be acted on. Can you talk a little bit about throughput in general? If someone wants to adopt an HSI and they’re an industrial environment, can we keep up now and how do we do it?

Christian Felsheim: [00:14:17] Yes we can. It certainly always depends on the application. It depends on how much you want to get out of it. How much spatial resolution you need. But if you ever stood beside one of these machines where you find these spectrometers. An example, let’s just take the plastic sorting, how these plastic elements fly over the belt and then are sorted and by air jets. It’s impressive, or we just had another application coming out of the seafood area, where these products are flying at 400 feet per minute over the belt and then getting sorted automatically. It’s fascinating. And as you mentioned before, it’s nice to tell the kids. They are impressed. But not only them.

Winn Hardin: [00:15:09] I think they’re impressed by the oyster. I don’t know about you, but conveying technology to my family can sometimes, dumbfounded discussions around the Thanksgiving dinner table, that’s for sure.

Mark Willingham: [00:15:20] I’d be remiss if I didn’t mention that this isn’t the future, right? We’re in the future now. We have well over a thousand hyperspectral imagers out there running in industry today. Not remote sensing but just in what we were calling machine vision. So this is heavily deployed across multiple sectors as we speak. As we sleep tonight, our imagers will be identifying either plastics that don’t belong or parts of seafood. In the seafood market, we don’t get specific, to try to make a better, more sustainable production of volume seafood and other things. But there’s a couple good examples.

Jimmy Carroll: [00:15:58] Yeah, Mark, that kind of ties into a question that I wanted to ask you guys about. And I’m also familiar with Headwall Photonics from a past life, when I was at Vision Systems Design magazine. And I know that historically you’ve been involved with projects with the EPA and NASA and the Department of Defense and even Ocean Spray. I remember that one, off the top of my head. But now not only with the idea of hyperspectral imaging almost being like, democratized, but in addition some of your acquisitions, like perClass, for example, what are some new applications? What are you opening the door to and what are some applications where maybe you haven’t seen your stuff deployed yet, but you could see it down the road in the next couple of years?

Christian Felsheim: [00:16:43] Maybe allow me to give you more background on why we made these two additions of perClass and inno-spec. You mentioned before what limits deployment of hyperspectral imaging. Usually in the past has been just a research tool. Now getting into a wider deployment, and two main hurdles are the complexity. You are getting a lot of information about your product, which you have to digest. And the second certainly is you need to make it robust and stable to work 24/7. And that’s exactly where these two acquisitions help us a lot. The perClass team, these are PhD machine learning experts, and they have worked to make analyzing hyperspectral imaging easy. Not less powerful. Actually, it’s one of the most powerful software in the market, allowing for object classification. Different in typical indices to being calculated automatically and also regression, so also doing quantitive analysis. But most of all, it makes it easy to use, easy to work with these complex data. This rich complex data, not for experts, not for spectroscopy experts, but for application experts which have not worked with spectroscopy before. We want to bring the technology to the application and not the application experts to the technology. So that’s what we can really do with perClass, and that’s why we took them on board. It’s really exciting to see that when you’re in front of customers and show the capabilities. And the other part is the robustness of the whole instrument, and that’s really where inno-spec comes from. As Mark alluded, they have like a thousand systems in this very dirty, difficult environment of plastic recycling. And these machines run 24/7, parts of them under very different climate conditions. Because these holes these machines are standing in, they are just open to the environment. And they run in winter as well as in summertime, 24/7. Very stable.

Winn Hardin: [00:19:02] Not sterile environments. Recycling centers are not known for being sterile environments.

Christian Felsheim: [00:19:05] Not at all. And this ease of integration and on the other hand side robustness and stability, that’s exactly what these two acquisitions brings to us.

Winn Hardin: [00:19:19] As part of that with perClass specifically in the case of, using bin picking as an example, bin picking is a difficult machine vision married to vision. Trying to pull a singulated object, which is usually loosely packaged, out of a confined space. So you’d see a lot of the VGR, vision-guided robotics people would have. Okay, this is for bin picking cylindrical objects. You’d have basically application based or pre-trained. In the case of AI, when they came in, pre-trained models that would be starting points for, say, food inspection versus plastic inspection versus an assembly type. So is how perClass or Headwall larger approaching the applications to make it more easily intuitive for the designer?

Christian Felsheim: [00:20:08] Well, the way the systems work, you always teach to the application to achieve best performance as well as in terms of efficiency as well as in terms of effectiveness. But this training is really easy. You just label your samples, you do some ground truthing and then label it in the image. And that’s the learning process that you do. And then you add to that over time, making your own library richer and richer. Yes, we do on the one hand side have a great experience in many applications. Like I said before, we did like 600 application studies across these three teams, perClass, inno-spec, and Headwall. But on the other hand side, we certainly also recognize that to many of our customers, this sorting knowledge is specific IP to them. So we are also open to really save that IP for their purposes. It all depends on what the needs and also the commercial needs of our customers are.

Winn Hardin: [00:21:15] Gotcha.

Mark Willingham: [00:21:18] One thing I would add, and that’s a great explanation there. You know, certainly we see hyperspectral in the market and continuing to come to market and not to replace RGB cameras or machine vision. Each has its use. Most common for us is when we start in a place in line that’s adding value. We immediately recognize with the customer other places in the line where value could also be added. So it’s much more of an expansion than any kind of technology replacement. It’s bringing new insights, and it’s at speed, which is really critical. What’s really exciting even beyond that, though, is we get so much more information, and if we’re working with a customer in one location, we’re also very likely working with similar customers in other locations. And historically, people would be very secretive or private with their information. But now we’re able to take these different but similar groups with this hugely rich dataset and elevate into a consortium where we apply additional AI to that, and we’re able to bring back insights that they could never find in their own one site. They’re only available in this type of setting. And where we have so many deployments in these verticals, as that grows, this gets even richer, right? So we see people changing from feeling secret about it to wanting the insights that can be provided by being part of this broader look. And this is something, again, that inflection point that AI is bringing those opportunities to do data science in addition to the sort of machine learning that we see going on more locally.

Jimmy Carroll: [00:23:09] Yeah, to me this is a really interesting intersection of technologies. And I’m just curious. I mean, we’ve kind of talked about it, but in terms of this new direction that Headwall is taking and the impact that you could have on the machine vision market, what else are you most excited about? Next couple of years? Do you have any any predictions or high hopes or anything like this?

Christian Felsheim: [00:23:32] To be honest, my most excitement comes from seeing the opportunities of these technologies and the value we can bring to our customers. And we currently really struggle to focus because there’s so much coming from everywhere to really say, hey, there is an opportunity here. And in most cases, we do indeed identify potential to really generate value for our customers with this technology because we are at an inflection point, as Mark said, where technology on hardware, software as well as the demand for more quality and saving resources come together. And that’s really exciting to see that.

Mark Willingham: [00:24:21] You mentioned like we can tell people that we bring food to market in a better way. We help our customers be more efficient. That lets them produce more volume and bring more, better produce to market. That’s what we really focus on. That’s our mission in life is not not hyperspectral data but improving the supply chains in some of these different verticals where we really add value. And again, I can’t stress enough that the moment is now. I think if you have listeners out there saying, hey, we’re not really using hyperspectral, I can tell you you’re behind the curve. Others are and opportunities exist. And we would love to hear from you, share with you where we are in that vertical, but work towards it. This is out there. It’s happening around you, if not within your own factories.

Jimmy Carroll: [00:25:13] One thing I just want to ask about that Mark, and I think that’s probably something that’s come up with you in the past and with customers or just new leads or whatever. But how do you go about, from an application analysis standpoint, how do you go about working with somebody to decide if they’re best off with RGB or multispectral or some implementation that uses whatever, filter wheels and things like that, or for your solution, how do you work with them to figure out if hyperspectral is best for them?

Mark Willingham: [00:25:44] A couple of different ways, and Christian I’d love to hear your thoughts on that great question as well. But, you know, one of the things that we offer now, we have this fantastic, essentially lab in a box that we can sell that lets someone get started with perClass software on it. As Christian mentioned, we’re not expecting you to know anything about hyperspectral imaging. You have to be an expert in nothing. We’ll help you with the range you might look in depending on your application. And then you can do some work on your own that we support remotely or with models that might preexist from other work, but it really lets people, that’s always been one of the challenges. And I think when people think hyperspectral, there’s this long period between initial touch and some kind of real activity that’s in scale. That is now very short because they can on their own really assess that and understand and see internally. We have customers that bring their own data to their board of directors to make an investment. We no longer have to go in and do that for them. And so we are populating the world with these stages. And they really help people get to work. They don’t need us until the volume questions come along. And that’s really where we add the most value is plugging it into their level of automation, their level of scale, and helping them realize it, not just learn about it.

Christian Felsheim: [00:27:07] And we are lucky enough by now that we really have these application teams. We have them in the US and the Netherlands as well as in Germany. And what we typically offer is to do a quick feasibility study right up in the beginning. That is pretty much a short-term effort at low investment, low cost. And then we can quickly say, yes, it’s a feasible technology. Or you might well do that also with a black-and-white or RGB camera or multispectral camera. As I said, we have so many applications coming for us. We have no problem in telling that right outright. And once we have done that, then we would go into a more detailed study where we really work with the customer to identify his needs. And the commercial needs are certainly a big part of that. How can we fit in? Do we achieve the value he needs for the investment? And then we start setting up a project together with him. And we have project managers driving that through this process, keeping in touch with the customers to ensure the success.

Winn Hardin: [00:28:23] Is there a short list. And this I think is going to be very application-specific, but it’s a list of primary criteria that come to mind that if a customer was looking at this problem on their shop floor, they should think HSI. I mean general automation is because usually all too often there’s jobs that can’t be filled. The existing manual workforce can’t keep up with productivity requirements. From HSI, is there a couple of things that come to mind?

Christian Felsheim: [00:28:52] I said that before today, we used to think about machine vision and identifying where is something.  Is there anything and where is it? That can perfectly be done with black and white. Now hyperspectral imaging tells you more about what is it, what is the characteristic of my object? What are the parameters that can be described for this? And whenever you really want to know more about it, when you want to grade a substance, when you want to differentiate it from similar-looking substances, then hyperspectral is a great opportunity. And as I said before, just contact us and we do a quick feasibility. And if it turns out it can be done in a cheaper manner, we will tell you.

Winn Hardin: [00:29:45] Mark, that seems tailor-made for organic products, right. Where we’ve got nature’s variability. But can you give us, I assume that this material science, this material detection capability is also probably important to other industrial applications. Are there some other ones? And I would love to talk about agtech all day in terms of pesticide, real-time monitoring, I think agtech is one of the markets of the future for all industrial automation and imaging systems. But more other industrial applications. Do they come to mind, Mark?

Mark Willingham: [00:30:18] Yeah, I mean sure. So we talked a little bit about recycled plastics. There’s also a very strong growing textile market. Kind of similar on the chemistry side so to speak. But also there’s so much that people can see when they’re producing electronics. So a great market is electrical components, and whether it’s soldering processes or other connectivity processes where we’re able to see quality in a way that a camera it would be invisible to. And certainly we talked a lot about food, and you can see there’s two hearts in your onion and things like that. There’s so many applications there. But then you’re kind of leaning towards the ag side, and we’re in that market as well in the remote sensing end, where we are able to get our hyperspectral solutions airborne and look for much more broadly, we have a very robust market in agtech around bringing hyperspectral data to farming to understand where you have disease, need fertilization, have a water issue. There’s just a myriad of issues that kind of marry very beautifully with once that goes to market, we can be there as well. And we’re doing more applications where we’re actually looking at both. But they don’t need to connect to both be independently valuable. And the other one I would mention that’s really growing in the commercial market and remote sensing is mining. There’s a huge amount of information that we can see with regard to how to more effectively do mining, so that’s a big growing sector for us as well in the remote sensing world.

Winn Hardin: [00:32:00] And it’s my understanding that the inno-spec acquisition actually is making your systems much more robust for hazardous environments. I mean, so we could probably envision a Headwall product to be actually in every harvester at some point, you know, where you can optimize it for each stage of the crop and for lots of different crops, even though many farms will focus on a particular.

Mark Willingham: [00:32:23] And here certainly also it plays in our favor, the core IP where Headwall is coming from. The core of every spectrometer is the grating that diffracts the light in the different wavelength components. And it’s very much about really matching the grating to the spectroscopy design to the application. And we make our own gratings, and Holographix is replicating them. So we are able to really build the whole system starting from the core of it, which is the grating to the spectroscopy, to the software, to the packaging with inno-spec, to purpose. And that’s a great advantage. And a big part of that is certainly also if you just look at our remote sensing products. Here it was essential that they are lightweight and really perfectly fitting in order to improve the flight time. And if you look, if you compare, an example, our spectroscopy systems, they are much more lightweight and more compact than others. And that certainly makes it easier to integrate, as you just outlined, into fertilizers or any agricultural machines. And they also have less mass to carry.

Winn Hardin: [00:33:40] So there’s a certain amount of customization for these systems, but they also can still have a very small footprint, use open data interfaces, so that they can interface with whatever traditional controllers are available and get it done. Don’t don’t need a lot of specialized computing hardware, even though we’re leveraging AI at this point. So all of that just opens up the door of opportunity to the application base, without question.

Mark Willingham: [00:34:09] And Christian mentioned our biggest challenge is to make sure we stay focused and pointed. And that’s part of what I’m trying to do here is to make sure that as a team across multiple continents and multiple teams, we have a common goal and we stay focused on it. And that’s where the challenge is. We’re not worried about competitors. There’s going to be so much market available that everybody can socialize the technology in a way that people understand. Just our own ability to bring it to market, and that’s really the focus of our teams. I will say, though, we’ve had these multiple acquisitions, we have more to come this year. Stay in touch with us. If you’re in the optical components market and you probably see some consolidation happening, and feel free to call and be part of what we’re doing here. But also on the hyperspectral side, the world is growing and we’re growing scale fastest in a way that really benefits the customer. And I think the teams have seamlessly worked together just amazingly well. And in every location where we’ve had an acquisition, we’ve grown the team locally. So this is a value play, not some kind of cost play. And it’s really on the move.

Winn Hardin: [00:35:25] Mark and Christian, there’s a lot obviously to be excited about right now at Headwall. And that’s coming through when we talk to you here. Beyond what we’ve already talked about, is there anything else you’d like to add before we finish up? I know we could probably sit here and talk for hours.

Winn Hardin: [00:35:41] Is there one thing coming down the pike, Mark, that in the next five, a couple of years, is there another holy grail approaching or are we just mature at this point?

Mark Willingham: [00:35:54] Oh, mature? I don’t know. I think the key point to take away is that there’s the inflection, right? All great inflections come from multiple technologies finding that at the same point. And we’re truly there. The edge computing that’s really coming out. NVIDI and other AI-focused processors bring the ability to do much more computing at the point. It’s perfect for hyperspectral, right? It’s just what was needed to enable it and why five years ago it looked so different, it took so long, and today it doesn’t.  And you combine that with the software, not just software but AI software, machine learning software, and data science to bring together the output of all of this. This is also coming out at the same time. So combine that with our history of harsh environments, which certainly remote sensing also is, but our also excellence in optical components. And our dedication to the craft is seen in our acquisition strategy, not to make a purely vertical but to have expertise in the high volume, replicated optical components that make the best hyperspectral imagers. So those all come together in the market, and you can see how we’re building something special right here at Headwall.

Jimmy Carroll: [00:37:15] Well, like Mark said, if anybody out there who’s listening or watching has any questions, they can reach out at HeadwallPhotonics.com. Or find these gentlemen on LinkedIn, and of course we’d be happy to take your questions and pass them along. Our website is manufacturing-matters.com. So Christian, thank you so much for the time. Really appreciate it. It’s been a pleasure.

Mark Willingham: [00:37:38] And thank you guys. Wonderful being associated with your podcast. And these are great questions, and hopefully it’s just the beginning. In a few months, things will be even different. So let’s talk again.

Winn Hardin: [00:37:50] Look forward to it. Thank you gentlemen. Have a great day. Anybody wants to see our old episodes go to manufacturing-matters.com, where you can find all of our past episodes. Be sure to like this and check us out on whatever your favorite podcast platform is. Until our next episode, have a great day!