Unaligned #10: Lumafield

This is a special treat. Who are you?

I'm Eduardo Torrejoba.

Start all videos for a lot of the reasons.

Yeah, I'm Eduardo Torrealba and I'm the co founder and CEO at Lumafield.

And what are you all working on?

So at Lumafield, we make hardware and software that brings artificial intelligence to manufacturing companies to give them more confidence about the things they make and help them make decisions significantly faster than what they do today. We do that with some really awesome technology that allows us to build computed tomography or CT scanners at a fraction of the cost of what they typically cost in the past.

All right, I've been in a CT scanner. It's a complex machine to make, Right. What role does AI play in making that machine? I'm happy to announce that I am now taking sponsors for unaligned. I'm proud to have AI Top Tools as our first official sponsor. One of the most important things for me running my podcast is staying on top of the industry, knowing all the players and what they can do. I made a complete list of companies in the AI industry on x.com but no one has as comprehensive a library of tools as AI Top Tools. They break down products by use case like productivity, media, chatbots, customer service and more. A lot of companies are looking for the exact right tools for their needs and there's no better place to find it than AI taught tools.

Yeah, so at Blumenfield we have, we've reinvented industrial computed tomography. So the CT scanner you've been in would be a medical scanner, I assume. Yeah, you've had a really experience in an industrial CT scanner. But they're machines that they're, they're designed to have much higher power than a medical scanner because they're trying to punch through metal and plastic and glass instead of just flesh and bone. And they typically have much higher resolution as well. Because inside of a person maybe you care about something about the size of a millimeter, but for industrial applications you might care about 10 microns way smaller. The way this all ties into artificial intelligence is that the ability for an AI to make a decision or make a recommendation or design something is totally contingent on the quality of data behind that model. And so the better the data you have to feed into these models to improve the decision making, the better the. It's garbage in, garbage out. Right. We want to have really high quality data for high quality decision making. And yeah, we've built an entire data pipeline and a machine to make this data available for our customers. In the, you know, plan of one day, giving them the ability to automate massive parts of their manufacturing operations.

Very cool. What are you learning by doing this? Because it's a new kind of manufacturing. Right. I've been in a lot of factory floors. I've never seen a CT machine.

Right, yeah. CT scanners are pretty rare.

I don't know, causing this. Sorry. Yeah. All right. So back. Yeah, sorry about that.

Go ahead.

So, yeah, so what's going on? So, yeah, tell me. Let's get back to that. Tell me a little bit about how is this changing manufacturing? Because I've been in a lot of factory lines before, from places that make baseball bats to places that make planes. I've never seen a CT machine on a manufacturing line.

What did.

Tell me a little bit about how the world of manufacturing is changing because.

Of what you're doing. Yeah, totally. So CT scanners today are really scientific instruments. They're designed to go into a special lab and they're. They cost up to a million dollars, sometimes millions of dollars, PhD, fully trained person to operate them. They're complicated, they're expensive, they're rare. We're turning it into an engineering tool. Right. A scalpel is a precision instrument that you only find in an operating room. But a pocket knife is something that everybody has. And so we are transforming this technology. We're making it less expensive, we're making it easier to use, we're making it faster to make decisions with. We're putting all the data in the cloud. We're transforming everything about the process so that we can take it from something that all these scientists have access to, something that every engineer uses on a daily basis.

Is it like cern? When I went to cern, where they have these big sensors and are studying particles colliding the first row of computers is about throwing away data and passing along higher quality data. Is that sort of what you're using the AI for? To see a better pattern in the data coming off this machine?

Yeah. So we use artificial intelligence in a number of different places throughout our data processing pipeline. One of the ways that CT scanners are hard to use today is that you have to set up the scan to get a good view of something. If you think back to a medical CT scanning situation, that would be like the technician, right? They like put you in the scanner, they tell you lay this way, put your arms here, put this thing. All the stuff they have to do, set it up. They have to set you up so they get a good scan. There's a lot of Complexity in that setup process, especially if you have metal and plastic together, or titanium or whatever it might be. And so we have created an automated system that leverages AI to make it much simpler. You just put an object in the system, makes all the decisions about how to optimize data acquisition. And then once we have that data, we are able to go from the raw data to the 3D reconstructions that you can view in the browser. Through our product Voyager, we do that in an automated fashion and we use AI in a few places there. And so that's the kind of thing, think of that as the, the radiologist, right? And behind the scenes, they get the data, they look at that, they turn into something, they use it to make decisions. We automate all of that so that the user that's an engineer, they don't have to be an expert in CT scanning, they don't have to be an expert in three dimensional reconstructions. They just get in and they're interacting with a 3D model just like it was a CAD model. And then on top of that, we can actually automate the decision making around the 3D model. So if we say, hey, we care about porosity, we care about alignment of the battery terminal, we care about cracks, we can automatically detect all of that using artificial intelligence and then make a go, no go decision in line during the manufacturing process. So we're really using artificial intelligence. And honestly, most of the stuff we're doing would have been possible three or four years ago. Right. It's not like we need the latest large language model to be able to make some of these decisions. But the quality of the data and the quantity of that data is the thing that we're changing dramatically for our customers.

So I've been on a manufacturing line in China that made hard drives and in the middle of the line was scanning electron microscope and they were sticking chips in. That is to do quality, find out why something is failing. Right. Is that sort of what it would be used for on a manufacturing line, to do quality control on a very precise.

This is all a quality control process. And so if you think about it, the example you gave there with a scanning electron microscope, that probably exists in a place where they run into an issue and they're trying to diagnose that issue. So we have lots of customers doing that today where, you know, standard production is just flowing, but as soon as a problem pops up, they can CT scan the object and they can see immediately if there's a problem with glue inside of a joint, inside of a Enclosure. Right. That's just invisible. Otherwise. We also have customers who actually use us to scan every single object that they make to verify that it's good before they ship it off to their end customer. Got it. So there's inline inspection and problem solving, troubleshooting workflows. We support both of those things.

How big is this machine and what kind of thing can be shoved into this? Could you put a whole cybertruck in there or would you just put the electric truck?

Or the scanner would survive in that situation.

Oh, you've got a magnet.

Not the magnets, just you would crush something. Right. Like the scanner is like the size of two refrigerators next to each other.

Got it.

Three feet deep, six feet wide, six feet tall. So we can scan something that is roughly the size of a basketball. And that's a lot of stuff. It's not everything. You can't put an entire car or an airplane, but you can put a lot of car parts in there. We have a lot of customers who make car parts, medical devices, athletic equipment. There are a bunch of constituent pieces that go into things that sit really well inside of our chairs. Yeah.

What does it look like to the operator? Are they talking to it or are they seeing reports on a screen? Hey, this part failed. You gotta go and fix the machine again? Something like that.

Yeah. It's just a touchscreen. You walk up to the scanner, open the door, put the part in, press a couple buttons to say run. Then you get a notification on your laptop or your phone. You bring up the data and you can look for whatever you want immediately. Or if we have the automated system in place, it's giving you a red light or a green light based on every part that you're scanning.

And tell me about the arc, the software architecture that's running this. Are you using a large language model or are you using smaller, more specialized models that, that are looking for patterns or something like that?

Yeah, sure. From the. Without going into like trade secrets, we have a number of different places where we're using more classic ML technology techniques or computer vision esque techniques throughout the pipe processing pipeline. And then once we get to the decision making portion of things, we use a number of different models in different places where we get the right level of leverage. If you go into our software Voyager and you click Atlas, that's the. What we're calling AI at Lumafield is Atlas. And you launch Atlas, you'll be chatting with it just like you would with ChatGPT. And so there is a large language model component to that where you can say, what's wrong with my part? And then it'll run a bunch of other analysis for you in an automated fashion. It'll actually generate bookmarks and views and recommendations. So we do have a generative AI portion of the product as well. But most of our customers, they're getting value out of that sort of like fundamental 3D volume that we create for them. They don't even know they're interacting with AI. Right. There's no part of it that they think of as I'm going to press D I F button, but the output that they get is using a lot of that processing. And I think that really is the future of artificial intelligence, is that it really goes into the background and you don't even realize that you're used to just the way that your Netflix Q or lots of other things are AI powered. You don't really think about it, but that's the reality of the world we're headed into.

What kind of fine tuning did you have to do the LLM to get it to understand what the engineer is trying to say? Because they're using language that's different than I would use with a machine.

Yeah. So we have some special techniques that we use to fine tune. We didn't train an LLM from scratch. Right. We didn't spend a billion dollars building our own large language model. But that's one place where the tools have come a long way in the past few years. And we're really able to in the same way that we didn't build our own data center and we didn't create all of JavaScript from scratch, we're able to leverage all of these other tools and plug them together in a way that's very different than most people do, especially in our space. Most manufacturing equipment and software is stuck in Windows 2000 World.

What was the challenge on building this machine, and particularly from the AI or software standpoint.

Right. Yeah. Yeah. So at Bloomfield, we've spent, you know, tens of millions of dollars on research and development to get to this point. We sometimes joke that we've had to build multiple startups worth of technology to get this product out the door. We had to build hardware, we had to build our own X ray detection technology. We have material scientists on staff. We've had to build an entire algorithms team to do reconstruction, to take the data from a scanner and turn it into these 3D objects that our customers can interact with. There's a whole pipeline of backend technology that we're using there. Then we have an entire web dev team, right. We have product people, we have front end developers, back end developers, DevOps people. We have all of this stuff that we need to communicate the data to our customers. So yeah, then the AI stuff, we've actually just homegrown all of that. We taught ourselves and trained our engineers on how to do that because we have spent a lot of money, but we haven't spent AI money on building and hiring devs that are making multiple hundreds of thousands to millions of dollars a year. We're not able to afford that as a small startup. And so we've had to really be a lot more efficient about the way that we approach this process because we're the people using the tools that those companies are creating and we're trying to bring that to our customers, many of whom have never even interacted with. I think probably the people that you and I interact with socially are totally. They have every app on their phone and all the AI models and let me check with Bard and then chat GPT and then a lot and all these things. But our customers, there's a whole bunch.

Right. And do you have some Chinese models running?

We do not. We're a very US centric company.

Right.

But no, our customers, they've never interacted with this stuff. Right. They're like a factory manager in Pennsylvania or something and they're just trying to get their shift. So we're giving them a tool that they've never even seen before. I think that's one of them most fun parts about building. What we're building is our. We want to make it so our customers don't have to care about the complexity or how hard of a technical lift it was to build. I don't walk into a factory floor and talk to somebody who's running a hundred person shift and brag about this clever algorithm. They don't care. It doesn't matter. They care about if they're getting the information they need to make a critical decision for their team so they can be successful with the injection molded parts that they're making for their customers.

I know a guy who makes little rocket engines mostly on a 3D printer system right out of manufacturer. Could I put this kind of part in the machine and know whether the printers fit it all out properly and is really precise?

Yeah, we've, we've scanned some rocket motors ourselves. I'll send you a link after this to an Estes model rocket motor that we scanned. We actually scanned a couple of them recently. We were talking to somebody making Something similar. And we had one of our engineers, like, went home and found SD's model rocket motor from the late 60s, and they had a modern one. And we scanned them and looked at them side by side. It's changed a lot in a few decades, I bet. Yeah.

So it can be used on all sorts of different kinds of manufacturing. Not just making electric motors for an electric truck or something like that, right?

Yeah. You could use it on medical device consumables. Right. It could be used on inhalers or epipens or running shoes or water bott or batteries, all kinds of stuff. That's one of the funnest parts about this job, is just like getting to see what our customers are doing with the technology and all the amazing factory tours that I get to go on to see how our machines are being used.

They're a lot of fun. Since you're doing that kind of work, what are you seeing happening on manufacturing lines? How are they changing? Because not just your technology, but how are they changing?

Right? Yeah, I think they're changing a lot. I think there are two, two or three kind of competing forces that are really pushing on factories in the United. This is all domestic, right here in the United States. There's a lot of different situations globally. One of those is a shrinking workforce of people that are interested in working in factories. I talked to a guy who runs a factory here in the northeast. I won't be too specific, but he told me that he has hundreds of employees that work in this factory. And when they hire somebody new, they give them a 90 day probationary period to be able to do their job. He said, if I hire somebody who's over 40, they always pass. If I hire somebody under 30, 90% of them wash out and quit because they don't want to do the work anymore. And I get that. I'm not. This is not me being like kids these days. Right. Like, I think that factory labor is just very different from the kind of information technology work that a lot of younger people want to do these days.

I ran a wave machine, a wave socket daughter machine at HPP when I was younger. I didn't want to do that job ever again. It's not a fun job compared to using your mind and talking to cool people around the world. Right?

Yeah. And so our, our customers are finding that there's. There's less interest in this. But at the same time, there's a huge demand to increase manufacturing output domestically because of what's happened in the last four years. Right. People have Woken up to the reality that fully globalized offshore just in time. Manufacturing is probably not strategically the right move for every business. Because if you're in a country where they can just turn off everything for two months at a time, maybe you can survive that if you're a Fortune 50 company. But if you're an SMB of some kind, they might have just doomed your business. And so they're trying to find ways to bring this stuff back so that they can have much more control over their own destiny.

I was in China when they closed all the factories in Shanghai for weeks because they were trying to get the air cleaning up for the leaders to come for the G5 summit, right? To have clean air for the leaders.

Yeah, exactly. So there are all these things. It's not even just Covid. It's just. There's all kinds of stuff where you just don't know if your factory is going to be able to produce something that's happening more frequently. And so there are. There's this decreasing amount of labor available, but this increasing demand for manufacturing and that has led to more automation. That was the. That was the way that everybody's solving this problem. I was at a factory a couple weeks ago where they were making this really simple, small, mechanical, purely mechanical assembly to go onto one of their things. And they used nine robot arms to pick and place all the little pieces to put it together. And it was something like you could just by yourself, put this together in 30 seconds, right? It really wouldn't have been that hard, but they couldn't find the labor to do that. So they have this manufacturing cell that's just like putting all this stuff together. Cameras, all these things. And then at the end of it, it spits it out. And it's like a design from the 70s, but they're still injection molding, but they're manufacturing it with robots. And so there's more of that happening. I wouldn't say that's everywhere, but there's just a lot more of this dependence on automation for assembly and then automation, the feedback loop for that process. And that's where Lumafield comes in. We give you that feedback loop. We allow you to automate things that today are still fully manual. That same factory, I was just talking about that, all the robot arms, it had a room with four people in it that were testing something the same way that you would test it at a. Like a car repair shop in the 30s to look into this very specific issue. So the technology disparity in factories is huge, but they Want to adopt technology, they want to be innovative, they want to lead. And that's one of the most exciting things about it. There's a lot to be bummed about. It would be great if there were. If we were making everything we made in the seventies in the United States still. But I think there is a wave coming. And I think I wouldn't bet against the people running manufacturing planes in the United States. They're pretty scrappy.

Yeah, I agree. I've visited a lot of them. They're cool people running big machines sometimes. Tell me a little bit about your competitors. What's the competitive landscape for this kind of machine? Because I assume this fairly expensive along sales cycles. You probably know who you're up against, right?

Yeah, we are. We're much less expensive than our competitors. Our. Our Systems are basically $75,000 a year. Hardware, software, servicing, support, maintenance, whatever you need, parts all included. So it's a fraction of what it would cost to get a traditional machine. The traditional CT scanner is 400,000, 600,000 million dollars for the machine. Then you've got a $50,000 software license and a maintenance contract. It adds up, right? There's the piece of this and these. These instruments are some of the most sophisticated pieces of technologies that humanity's ever built. I'm really not overstating that. Right? Like, it's the same kind of stuff. The brands that make these systems, companies like Zeiss and Nikon, right, they are making parts and systems that go into semiconductor conductor fabrication facilities. They're just like the bleeding edge of what we know how to do as a species around technology. So they make these beautiful, amazing, incredible instruments. But it's like a hypercar Ferrari, right? It was just like you. You don't need to drive that down to the corner store to get a gallon of milk. And our customers, they need to go get a lot of milk, and they can't afford a fleet of Ferraris to do that. And yeah, our competitors, we very rarely end up in a situation where we are in some kind of shootout. 85, 90% of our customers have never had a CT scanner before. And they've never even thought about using it until they find out about Lumafield. So we really see ourselves as broadening the market for a totally new use case, more so than trying to compete against phenomenal company like Zeiss or Waygate or lots of the other ones. I have a lot of respect for those companies. They just do something very different.

All right, you explain how AI is getting on the manufacturing floor in Your instrument. Are you seeing signs that AI is running the entire factory floor yet? And if you don't, we all know that's coming. What's your advice to other AI entrepreneurs who are going after factories? How do they get customers? How do they get their technology? And what's the challenge of working with people who don't really know much about AI yet?

I haven't seen any signs that AI is running anything meaningful inside of a factory right now. I think it's much more like what I was talking about when there are these critical points where something is happening with a software package and that's using artificial intelligence, but the people involved might not even think of it that way. And it's usually a subset of a subset. It's not gonna be what are we making today right there. There's no like central planning AI that's doing this kind of work for any of our customers. Maybe somebody's doing that, but I don't know who they are unless they're doing it. Yeah, I think, I don't know that. I have a lot of advice. I've never been one to take a lot of advice and so I'm not great at dispensing it. I think I would tell people, go get out there in the field and talk to your customers. Manufacturing is a face to face business. It does not happen over a webcam. And your customers are not people who are working from home. Like, by definition, you cannot have a factory in your house. Like zoning laws, there's nothing from happening. So you need to go to the place where this is happening and talk to these people and understand the exact problems that they are encountering. Because the difference between what you're trying to sell them or trying to build for them and what they really need, it could be an inch or a mile, but it doesn't matter. As long as there's a gap, they're not going to use that technology. And if it's an itch, you can close that really quickly and you could end up with a great use case. You're never going to find that out from your office or from your house. I try to spend at least one day a week the customer facility somewhere in the country. And it's. They're wildly different there. There's some things that are similar, but manufacturing is the ultimate long tail. Every object that's produced is manufactured. And the way you manufacture it looks different in all of those places. So trying to build a general solution for manufacturing and for AI and manufacturing, I don't think that's going to work. You need to go talk to individual customers and solve their problems. It's the only way to do it. It's pretty old school, so I think it's what people have.

Well, it sounds like you're doing it, which is awesome. What else are you learning by being in the factories and building a very specialized piece of equipment using AI? Any other things you're noticing happening in the factory floor that you know that you could get other entrepreneurs to key in on?

That's an interesting question. I think the. I think there's a real lack of a coordinated strategy for technology side of factories. You find people who are using computer. Literally a Sun Microsystems workstation I saw in a factory the other day with PLCs from the 80s.

That's old.

Yeah, yeah. And then you have 2D vision system running edge AI from Cognex. And those like they're not actually interacting, but they're making this. You know, you're making something with one system and verifying another one. And then the feedback is like a person turning a dip switch. So I think, you know, if you have any kind of preconceived notions about the sophistication or level of technology adoption inside of factories are probably wrong in either direction. Right. If you're thinking it's all dip switches and manual movements, no, there's like edge AI being deployed. But if you think because they have that, they're ready for this plugin that I have or this like centralized software system or something that runs on an iPad, that might all be impossible. And so I think anytime you're dealing with an environment where generations of technology that are being deployed are four or five generations apart within 10ft of each other, I think there's really interesting opportunities there to build something that is so much better than what they're already using that it becomes a no brainer to make that decision. That's what we're trying to do at Lumenfield. Right. We're trying to build something that gives you exponentially more information and better decision making than anything you could possibly use otherwise. I think a lot of people build, they take the productivity apps that they've built for other environments, like they want to be the notion for manufacturing. And I'm not throwing shade at anybody. I know people working on that problem, but that's a really specific thing that most people have no idea what that even means inside of the manufacturing world. So I think I've. Yeah, I think I'm always surprised by the huge contrast between the like tech tree advancement, Maybe like a sci fi term between different parts of every factory.

What does this look like in five years? Because we're getting humanoid robots in the next five years, right? And we're certainly seeing a lot of robots in manufacturing, but they're special. They're mostly specialized. How do you think manufacturing is about to change? Both because of AI running robots, but also because of technical tools like yours that are letting people do new things.

I'm curious why you think we're getting humanoid robots in five years.

Because I know some people at Tesla and their robot is learning fast.

Okay, good to hear. Yeah. I haven't seen any evidence of that on the floor, so maybe it's going to be one of those things that doesn't happen until it happens all at once.

Might be 10. I might be wrong about dates.

Sure. Yeah.

I'm always wrong about dates. I'm like, elon, dates and me are like.

Which is fine. I just try to be right about dates for my customers. I don't know where this goes in five years. I think I can see there's always an S curve for technology, right? Where, like, nothing's happening. Oh, my gosh. And then it's cool we're there, right? Like, think about hybrid cars is a great example. No, Hybrid cars. And then now there's tons of them, and they're not as good as they're going to get. I don't know where we are on that curve with artificial intelligence. Part of me thinks we're right up here, right, and we're flattening off. And we've. The last few years have been the crazy thing. But maybe we're right here, right? Maybe there's a long way to run when it comes to humanoid robots. I'm a mechanical engineer. I get, like, motors and torque and there's a bunch of unsolved problems. Boston Dynamics is right here. I have friends whose kids play with my kid that work there. I understand some of the things that go on inside of these places, and maybe there's a breakthrough that I don't see. But I think probably if I'm trying to think about the future, I really only try to do that with respect to the things that I have control over. I. I don't have any control over any humanoid robotics companies. And you're not Eli. I don't think anyone would ever accuse me of that. For better or worse, probably worse for all the people that own equity in my company. But yeah, that's not. It's not a thing I don't know, I guess I just, I'm not good at making predictions. Maybe I'd be a more interesting person if I was. But I can tell you that a year from now there are going to be significantly more facilities that have opened up in the United States because of federal and local tax incentives to build factories. We're going to have higher wages and manufacturing than we have today because there will not be an increase in the number of workers. Those factories will be understaffed and they will not be putting out as many things as the people who built them want. And as the tax incentives wear off, there will need to be an adoption of more technology in order to fully staff those things. And I think that is a place where maybe the need will be so great that the robots don't actually have to be better than they are now. They'll just be good enough as is so they can be used at a level that makes sense. I think the real kicker for all of this is what happens on a geopolitical level with supply chains writ large. Right. If, if Covid is a, every five year event, you're going to see a lot more robots in American factories than you see right now. Just out of necessity because there will not be a globalized world of goods being shipped in the same way that, you know, same way that everything was turned off in 2020. If that's a rolling blackout situation, nobody's going to put up with it. We're going to have to have more domestic production and it's going to have to be done with robots. Those could be robot armed, those could be humanoid, that could be anything. But I think the next 12 to 18 months will further expose the lack of manufacturing capacity that we have in the United States. And there will be more and more of a push to adopt either cutting edge or just existing technology for automation. So Maybe that's my 18 month prediction.

Sounds like 5 years market coming for people like you who are building new devices and new technologies to run the robots and check on their work. Right?

Yeah, exactly. That's, that's the bet we're making at Loyalfield is that there is going to be a need for more control and more feedback over these complex systems. And there's, it's not going down. Over time, the complexity of the things in my life has increased and that means there's more complexity in manufacturing and.

Customers like consistent quality. Right. And your device, I expect once, once you put it on a manufacturing line, the quality of what's going through that line should go up. Right?

Yeah, we really, we strive to really like constrain the bands of what can happen with something once you're inspecting it. And this has been the idea forever. Right. The entire process of Six Sigma that people go through is about getting to standard deviations are 1 in 10 million or a million or, you know, whatever the number went process they're running. We want to give people the tools to know if that's happening. And that's. You're absolutely right. We're. We're banking on more manufacturing happening and we're banking on the need for more sophisticated inspection regardless of the coming waves de globalization. It doesn't matter. There's still a lot of sophisticated stuff being manufactured here and abroad, and they can all use our technology.

If I was selling a widget to Elon, like a headlight for his new truck or something like that, I'd be bragging that I bought one of your machines. Check out my new CT scanner with AI in it.

Yeah, no, we have, we have a lot of customers in the automotive space and I think the ability to collaborate on the data really makes a big difference in the industry. Yeah. But I won't say that hasn't happened.

If I'm making an electric motor, I'd probably want to validate that motor past your inspection. Right. And probably there's some data on top of the motor when it goes to a car company or some. Somewhere else that they can look at. Oh, how was this made? How was it validated?

Yeah, exactly. The ability to say, hey, this is in direct compliance with the drawings and the specs that we agreed on is huge. To talk about automotive, maybe to take a deep dive into that for a second. And this isn't what you're talking about, Tesla. This is any car company.

Yeah. There's a lot of them making electric vehicles now. So they need a motor about this size that has a lot of complexity.

Battery cells and electronics and everything else. If you look at a lot of the largest automotive recalls that have happened in the last several years, they all have one thing in common, and that is that the brand damage is usually to the oem. Right. Whether that's. I won't name them. You all know car OEMs are. Right. Yeah. They're the ones that people think of as like, this company had a recall. The financial burden of that is actually borne by the supplier. So the supplier has to pay for all the service visits, all the new parts, all the replacements, all those things. So you might have a $4 billion recall and you've got a Tier 1 OEM that's cutting the check for that. And so they really don't want that to happen. Right. They want to have as much ability as possible to say, hey, we delivered everything exactly to spec, not our fault. So there's an element of covering your butt a little bit that's happening in industry, but there's also an element of, hey, do we all agree? Right. Is this exactly what we need to build? If we've tested it and it works, is every part that we're getting in compliance with the thing that we tested? So, yeah, there's a lot of incentive all around to make sure that these things are working correctly. As a user, I want everything I own to be CT scanned because I don't want anything to break.

Totally agree. We want a higher quality product for a cheaper price as consumers, and it's people like you who are making that happen. So thank you.

Yeah, I think it's mostly our customers that are actually making that happen. I think we just try to.

Oh, if they don't have one of your machines, they can't take pictures of their electric motors before they go to some car company. By the way, does your machine test every product going through a manufacturing line or just one out of a hundred, or is it fast enough to keep up with a high flow line?

Yeah, I mean, I get to do the ultimate engineer answer and say it depends. It really does, whether the object needs that. We have some customers who do like first, middle and last of a shift to make sure the process is working as intended. Other customers do a hundred percent, thousands of objects per week through one of our scanners. Some people only check something if a problem shows up. So it really varies across multiple different industries. The company that's ch, that's inspecting everything, they're making a car part. Nobody's CT scanning a hundred percent of car parts anywhere else. We're enabling the first version of that. We have another customer that's doing batch sampling. They're making batteries for electric vehicles. That's not a technology they had before they started using the field. So they're going into things that are changing the way that the world works. And that's some of the most exciting work we get to do is like empowering these people who are changing transportation or drug delivery or the environment or sustainability. Getting to work with those companies and make it possible for them to do something that they couldn't do before or do it faster. It matters. It's a really fulfilling job.

Very cool. Could your software be used on other Kinds of machines other than CT scan scanners?

Yeah, for sure. Yeah. Our, our process and the technology that we've developed around manufacturing data is very scalable. And we can actually take in 3D scans from any kind of industrial 3D scanner as well. You can upload those into our software and use our software to analyze that as well. But we're, we're very focused on just building CT scanning hardware and then having people bring other data from other sources to our software platform.

Very cool. Tell me a little bit about your company. How was it funded? How did you get the funding to build it? Because you said you spent, what, $10 million on R& D and stuff like that.

Oh, I wish it was only $10 million. That would have been great.

It's an expensive bill, huh?

Expensive. It's an expensive thing to build. Yeah. So we started the company myself and my co founders, we started in the fall of 2019. So just about four years ago, our seed round was led by Kleiner Perkins and Lux Capital. And then in the fall of 2020, we raised a Series A and that was led by dcbc. And then last August, we raised our Series B and that was led by Spark. So we've raised $67 million in change for the business so far and we've spent a lot of it. So that's kind of where it's going to be.

Probably raising some more money then.

Oh, if somebody wants to give me a good price, I'm not going to take our.

The price is the problem right now. The Fed made money more expensive.

Very true. Yeah, yeah. I don't. Again, I try not to worry about things I can't control, but fundraising has been interesting journey for a lot of people over the past few years.

Yeah, it's a tough one for people. I know that. Congrats on getting this done and thanks for helping American manufacturing because it's a big deal.

Yeah, right. It's a really fulfilling thing to get to work on. I'm not a politician, I'm not an economist, interested engineer. And getting to work on something that matters and has a chance to really make a big difference is a good reason to get out of bed every day, come to work and get to do the hard stuff that we do for our customers.

What kind of people do you want to have as customers and how do they find you?

Yeah, our customers can find us. Just go to lumafield.com we work with everybody. We don't have any kind of thresholds or expectations of small startups with five or six employees that have our machines. We have Fortune 10 companies that have our machines. If you're making a physical product and you want to make it better, we want to help you do that. And we priced our system in a way that makes that possible for people making everything from lipstick to medical implants. So we, we love working with customers across a very wide range of industries, and we'd love to talk to anybody who is interested in us.

Thank you so much.

Thank you.