[00:00:00] Speaker A: Man, we use a lot of stuff. We need a lot of stuff, and someone's got to make all that stuff. And it's crazy to me still that there's 300 person factories out there that are making the tiniest little part that you don't even know is on your Toyota Tacoma. It's just on there that's employing a community, and there's so much engineering and brain power that goes into building.
[00:00:18] Speaker B: That's.
Hello, everyone, and welcome to the robot industry podcast. We're glad you're here, and thanks for subscribing. My guest for this episode is Matt Jones. Matt Jones is VP of us sales and operations for Microsoft. And Matt and I met recently at the automate show in Detroit. Welcome to the podcast.
[00:00:44] Speaker A: Thanks, Jim. Yeah, excited to be here and talk a little bit about micropsy.
[00:00:48] Speaker B: Well, I wanted to start out with the kind of the typical question I ask people and reading bios and this kind of thing, but I want to find out how did you get involved and interested in robotics and AI and everything in between?
[00:01:00] Speaker A: About seven or eight years ago, I heard of this company called Kianz, a big automation supplier that many of your listeners, I'm sure, are familiar with and started working there. So they're really in the automation space. And really my first introduction to manufacturing, and to be honest, was kind of just fascinated by it. I mean, walking into a plant for the first time, or even still today, is just really cool to me, for lack of better words of man, we use a lot of stuff, we need a lot of stuff, and someone's got to make all that stuff. And it's crazy to me still that there's 300 person factories out there that are making the tiniest little part that you don't even know is on your Toyota Tacoma. It's just on there, and that's employing a community. And there's so much engineering and brain power that goes into building that. So from there, I was just really interested in that manufacturing space. And after about five or six years at key ends, mostly out of Nashville, Tennessee, and then transferred out here to the west coast in San Francisco, where I'm based, know had a lot of learning opportunities there, really got to know the industry, a variety of topics, and then a recruiter reached out, to be honest, and said, hey, there's this company called Micropsy, and they're looking for someone to kind of kick off their us team. So without too much of a backstory, Micropsy is a german based company founded about nine years ago out of Berlin, Germany. They were looking for someone that had a little bit of a sales background, but maybe also maybe a managerial mind to be able to build up a team. So that's where I started, and basically Microsofty pitched their product to me on the phone and I said, if this does what you say it does, this could be huge, and I want to join and be a part of this. So that's from there. About three years ago now, I joined Microsoft and have built our team from myself sitting in a tiny little office downtown San Francisco to a decent sized office. I'm here now with ten or eleven of us based here.
[00:02:51] Speaker B: Well, thank you for that. Congratulations, too. Now, micropsy, is it a hardware company, a software company? Where do you fit?
[00:02:59] Speaker A: Yeah, good question. We ask ourselves that sometimes. We are definitely a software company. At our core, the only thing we're manufacturing or developing on our side is the software. So AI software, to be specific. I'm sure we'll get a little more into that soon. AI, the buzword of the year, I would say. But our system goes on robots, and for that to work, there's a lot of hardware involved. So we have to know the hardware as well. We have to have robotic engineers, automation engineers that ideally have been in manufacturing plants that now work for us, that know the industry. So we find ourselves being a bit of a mix. But again, at our core, Microsy is.
[00:03:34] Speaker B: Certainly just a software company, and that's great to know. So when you talk about robots, I know I've seen the demo at automate and you had a collaborative robot. Are you collaborative? Are you industrial robots, or do you care?
[00:03:48] Speaker A: Definitely don't care. Traditionally, when Microsoft first started, we worked exclusively with universal robots for those first few years. By the time I joined, we worked with an industrial as well, and very exciting news. And what you would have seen at automate was on an LR mate, on a fannick, and that was a huge move for us. Fannick is huge. It's hard to grow a robot software company if you're not working with one of the biggest players. So we certainly work with universal robots still, all of their models excited for their UR 20 and to be able to work with that. But also there's plenty of applications where the payload is too high, the speed is too fast, and we also work with Fannick in particular Cuca, and continue growing that portfolio.
[00:04:31] Speaker B: And so this is going to be kind of sounds like I'm pitching for you and I'm not. But when you showed me the demo in your booth, it was like ridiculously simple. It was like a five minute demo. And I know we're here in a podcast, so we don't have visuals, but can you kind of walk my audience through that?
[00:04:49] Speaker A: Sure. Definitely. So, Mirai, our product, our staple product is a robot control system. What that means is normally when you're programming a robot, you're giving them XYZ coordinates. You're sending it to the same place over and over, as many of your listeners know all their ways to program using cameras or 3d technology. But what sets us apart, I would say, is we're truly using real time control. So we have a camera typically attached to the robot itself on the wrist, and it's almost like eyes for a human. It's a little cheesy to do the hand eye coordination analogy, but it really fits us quite well here. The hands being the robot, the eyes being the camera on there. But then you need a brain that tells it what to do. And that's really what our software is. It's telling the robot, okay, based on this image that I'm seeing in live time on the robot, I need to adjust to the left, adjust to the right, rotate on a z axis, et cetera, et cetera, and we're able to guide it. So what you saw there was trained in about 2 hours or so worth of effort, meaning, and by training, we say that specifically, it's not programming. You're actually grabbing a robot by the wrist or on an industrial. You're going to jog it and show it what you want it to do. Show it the area. There's no CAD files.
There's not a lot of pre work that you're used to on vision system. There's no real calibration. You're truly just showing it almost like you would show at a human what you want to do. We're able to take an enormous amount of data from that. All those recordings, maybe an hour worth of recordings. We crunch that, we're using our neural network and kind of making that local, and we save it, and we're able to run it in real time on the robot. So what you would have seen at automate is we were plugging a flexible rubber piece into an airplane wing. This is a customer we work with called Clickbond out of Nevada, who we've co branded or co marketed with a little bit that is an aerospace company, and we were able to train that, show it. And even if we're moving, we were inserting a rubber piece onto a wing. Even if we move that wing, even if that rubber piece is super flexible, which in this case it was. We're still adjusting because our robot, or our system, I should say, is talking to the robot 20 times a second and updating its path, updating where it needs to go, just based off that live image. So that live image, that real time control, as we call it, is really what sets us apart.
[00:07:05] Speaker B: So what are some of the other use cases that you're seeing? And I know I went to your website and checked out leak testing and screwdriving and gear picking and assembly and end of line testing. What are kind of the things that you're seeing more and more in industry?
[00:07:19] Speaker A: Yeah, I mean, kind of, as you just listed five or six there. That's almost what we're trying to figure out to some degree is there's so many different use cases. A lot of times what we're doing is we're rolling into a factory, we're rolling into engineers, or obviously at a trade show, and we're talking to dozens of people, and they're all kind of going different directions with it. So what we've kind of learned is instead of pushing a vertical and saying, hey, we're really good at screwdriving, let's just work on screwdriving. We'll try to present it to a customer. And often they're the ones that know their product, that know their processes. They know the applications that they want to put a robot on, but they can't quite because of some sort of variance, which, again, another keyword for Microsoft C is dealing with variances. So we kind of let them bring them to us. But to still give you at least a decent answer here, screwdriving is one. For example, there's an automotive company I'm working with here locally that's one of our biggest projects. That is finding a bolt on a seat, and it's simply screwing it in. Now, that seat comes down to conveyor belt. It's extremely vibratory. It's not incredibly consistent. It's actually hand fed into the spot. So we don't have an even decent XYZ coordinate. Seats are different colors, material, et cetera, et cetera. So we're able to find that bolt line, the screwdriver, up on a universal robot in this case, and screw that in. So screwdriving is a good one. Anything with cables is a huge highlight for us. If you all. Anyone listening, or even you, Jim, at the next trade show, there's a good chance we'll have a cable plugging demo going. Those always catch a lot of eyes, but grabbing a cable out of space, all six degrees of freedom and just plucking that cable and then plugging it in. So what does that mean? Electronic industry, server assembly or server maintenance even.
And then of course, again, automotive, there's countless cables. What you think? Big cables typically, but also small little wires, including ribbon cables, something we work with quite well. So any sort of assembly tasks, cables is a good example of how we can pick in place or assemble.
[00:09:15] Speaker B: And we kind of spoke about this early, just before we started, about cameras. And you're fairly camera agnostic, right? Like everybody likes to have maybe their own specific camera, but you don't care, do you?
[00:09:27] Speaker A: So we don't care in the sense that the camera is not that important.
We typically are using just a standard USB 2d rgb camera.
People often ask the specs, just like a good engineer does, understanding how many megapixels, this and that. And they usually are surprised, to be honest, at how low resolution that we're actually doing quite impressive, precise tasks that we do. So all that to say, I wouldn't call it camera agnostic in the sense that we do need to vet the cameras and make sure they're going to communicate fast enough. That's typically the key for us. Less than quality. And when someone purchases our software, they are getting a box in the mail. They're going to get the cameras themselves that we prepackage. Again, we have some options, but to be honest, the options are more for fringe cases or to be honest, supply chain. It's a business decision in some ways. If this one camera manufacturer that we use decides to have some supply chain problems or forced into supply chain problems, we want to be ready. But at the end we are much more. We're relatively agnostic, if that helps.
[00:10:33] Speaker B: So I wanted to also talk about lighting for a second because that's a really big challenge in all kinds of different factories. And here we were at a trade show where the lighting is always terrible. It's very consistent.
And I think you were turning off lights right during our short demo.
[00:10:49] Speaker A: Definitely. Lighting is a pretty cool one. That we can work with it again sets us apart a bit.
It's because of the real world problem. Lighting changes throughout a day. And in many factories, some, maybe there's less ambient light than others, but even still, you change out those traditional light bulbs with some leds and all of a sudden you're having to redo all of your cameras. A light bulb goes out and you got to fix it before you can even run a robot. That's no good. So what's cool is we're doing, as I mentioned before, imitation, learning a type of AI, and essentially you're going to show imitation, learning models, different variables, different possibilities. So lighting being an obvious variable, we can kind of play with different ones. So there's a big window behind me, and we actually recently bought some special blinds, like blackout blinds, so that when we train our demos or work on proof of concepts, we'll lower those blinds, record a few pieces of data, raise the blinds, open them halfway, and just kind of show different examples. Because when we train here in usually sunny San Francisco, with that light coming in, and then we go into a big assembly hall in Detroit where there's not a lot of ambient light, we don't really know what exactly it's going to look like. So we try to show it a variety, and we recommend our customers do the same thing. If they have a lot of ambient light, especially do some data recording in the morning, in the afternoon, and then ideally, if they're running three shifts, do a few on third shift. Because those nuanced light changes, while they may not make a huge difference on ours, it's just a great way to make it more robust.
[00:12:15] Speaker B: Great. And I wanted to kind of ask you, in your experience now, who's your perfect bullseye customer? Or is it an integrator, or is it a mix of both? Maybe we'll talk about the end customer first.
[00:12:27] Speaker A: Sure. On the end customer side, robot users are typically who we end up selling to. I mean, we have definitely been involved on a handful of projects where it's their first robot and they're getting involved with it. Typically, if they know the robot pretty well, it makes the learning curve on our system a whole lot easier. If they're an expert on universal robots, for example, and they know polyscope like the back of their hand.
Our system is a ur cap on ur. It's really just you're adding a simple device, and they need to learn our system, of course, and that's something we take very seriously. That's probably been our biggest learning this year, is getting customers properly onboarded. And as long as they're just learning that piece of it, it makes things a whole lot smoother. So, end user wise, anyone that is using robots, ideally, assembly, I would say, is probably the most common application. Working directly with engineers, typically, or engineering managers that are either. There's really two options that can go. One, they already have a robot that isn't quite living up to the job. They didn't realize when they put this machine in that hey, there's a lot more variance here, whether it's positional or lighting or even like, form or color, that there's more variance than we realize, and the robot just isn't working. Those are the easiest ones for us because you go in and you slap on our relatively low cost solution, and all of a sudden you have a $200,000 machine that wasn't working so well before, and now we've cleaned it up and it's working consistently. And then the other one, of course, like many automation companies, we're going after new lines or new cells, new products there. So what we've learned is we try to get one in the door, and then people get pretty excited. It's a lot of have to see it to believe it, which is why trade shows are so good for us, because they kind of start catching on. And it's really cool to watch another great engineer I work with here in the Bay area that they bought one for their large automotive company and then quickly realized, hey, there's like 20 other groups that could use this, and they kind of let it spread from there. So a young maybe I'll actually axe that from the podcast.
So, yeah, I think an engineer that has a lot of ambition and innovation and thinks forward is the best ones for us because they often bring the application straight to us.
[00:14:41] Speaker B: So I buy this unit from you. Maybe I'm not as experienced as I thought I was. What does onboarding look like and what does support look like?
[00:14:49] Speaker A: Yeah, that's been maybe a growing pain of Microsoft over the last three years is our system is relatively easy to use. I don't necessarily want to completely backtrack from that. At the same time, we went a little bit off of man, it's so easy to use. Let me hand you. We'll ship it in the mail. It'll show up and read the quick start guide. We'll jump on a virtual call and help you through it. And that worked for some customers. The customers that read the whole manual, the customers that have plenty of time, it's going into a lab setting, like they're just going to test with it. R and D, something like that. That mostly works quite well. But then they liked it in rnd and they put it into production, or started to put it into production. And that's where we started having some problems, where we just learned we really need to be there and we need to kind of hold their hand, really, through the application. So we've switched our model a little bit where when someone purchases our system now we essentially force them to allow us to come on site with an engineer, sometimes usually two contacts. So someone from engineering for sure, possibly their salesperson or possibly myself, and we'll go in and spend two, sometimes three, depending on how big of a project it is with them, where we're talking about every little detail, we're truly holding their hand for 16 to 24 hours of time, showing them how it's set. And we really hold customers hands through that first application. Most applications that we sell, they're purchasing one to start, but they have seven or eight lines that are doing the exact same thing. So we know from a business standpoint that if we put the effort in and make that first one work right, that 2nd, 3rd, 4th will come through. So even though it's quite expensive for us, at no additional cost to the customer to send someone for two or three days when they may or may not be local to any of our people, it's paid off really well. And I'd say we've been kind of implementing that for the last year and that's been a game changer for us.
[00:16:34] Speaker B: And on your cameras, you've got no camera calibration, right? Because this is done automatically, correct.
[00:16:40] Speaker A: I mean, there is a lens that goes on and you've got two settings on there. You've got your aperture and you've got your focus. And sure, you're playing with those a little bit initially when you're setting them up, but besides that, it's plug and play. And again, that's a 15 2nd job. You're looking at the live camera and you're rotating this little ring and making sure you're in focus. Again, we help with that quite a bit. Just make sure the camera image is incredibly important. I mean, that's all of our data. So we need that camera image to be good for what we're needing, I should say. So with that, that's really all. There's no backend camera software. All the camera interaction happens through our software. So you're not having to upload or download another third party camera system. It's quite easy. And that's something that gets people that know vision well. And again, at Kianz, I worked vision all the time and they make a great product. We made a really great product, but there was a lot that had to go into it. There needed to be training. The nice thing about ours is you need to know Mirai, you need to know our software, you don't need to learn a new camera. And it makes it quite easy.
[00:17:42] Speaker B: And so how do you get updates?
[00:17:44] Speaker A: Like whenever the product's update. Like a product? Sure, sure. So updates are huge for us. It's something that's really cool about being a software company is every two, three months we're coming out with something new, and that's something that we do pass on to customers without an additional cost. So we have a couple of updates that we actually just are finishing up right now. I probably shouldn't mention them specifically, but I will say a recent update that we had about six months ago, we were able to cut our cloud processing time down dramatically. So it used to be you would do your recordings, you would send it to a cloud and it would process, and that would take three, four, to be honest, when I started it was about 16 hours, which is painful. We got it down to 45 minutes, or about 40 minutes to be exact, and it should shrink even more. We were hoping the long term goal is to get that down to five minutes so you're able to process all of this data, which again, you're talking hours of video footage, typically extremely quickly, and you're running a fully functional neural network.
[00:18:45] Speaker B: That's crazy. That's great. Congratulations on that new update.
[00:18:49] Speaker A: And I guess I can finish on the update thought. Sure, yeah. But as far as when customers are getting an update, we usually are doing those over the air. So they essentially go onto their tablet that they use to train our system on, and they give us access to. It's essentially opening a VPN where we can go into it. That also is probably a good point to say for us to have any access to their data or the outside world, to have any access to anything in there, they have to check a box and it only expires after 24 hours. So from a privacy standpoint, all of our system can be run completely offline. There's not going to be a live camera recording things in their factory. Most of our customers unplug it from the Internet when it's in production because we have no reason to be. But if they want an update, they plug in the Internet, they give us access, and we can do an over the air update. If that doesn't work, there's certainly a lot of factories that don't want Wifi signals or anything like that going on in their factory. We can also send a USB or send a file and let them put it on a usb and upload it just directly on there. Whenever we have a software update, which, like I said, every three to six months, we have a pretty big release, typically.
[00:19:52] Speaker B: Well, that's great. Well, hey, Matt, thanks for coming on today. Have we forgotten anything to talk about?
[00:19:57] Speaker A: No, I think that pretty much sums it up. I can give out some of my contact information if people would like to learn more about us. You can certainly just find us at Microsoft Industries.com and I'm sure you'll share that as well. But otherwise, no, it's been great.
[00:20:10] Speaker B: I will put that in the show notes for sure. And also, I just wanted to get a feel for if I do have listeners who are interested to say I got to find out more, I've got to connect up with Matt.
What's the best way to learn more about the product and maybe getting the product into the factory?
[00:20:25] Speaker A: Sure. I mean, LinkedIn is an easy one. Matt Jones is not the easiest search on the Internet. There's a few of us out there, but Matt Jones, Microsoft see should pull me right up. Feel free to connect there. I'll give you my contact information for you to post with this link as well. And just reaching out directly by phone or email is great. And one thing that's really cool about Microsoft, part of our sales model is all of our salespeople spread out throughout the US have small robots fit in the back of the car, the smallest robot ur makes and pop it out of the trunk and can show you a demo right there on site. And hey, if you even have parts, hand us the parts and we'll have it trained and show some sort of example with it in less than an hour. So would love to get out to some more factories and show what we've got.
[00:21:09] Speaker B: Well, I tell you what, I was very impressed when I had the demo at automate show. Hey, when you're not automating and talking about vision and AI and such, what do you like to do? Do you have any hobbies?
[00:21:20] Speaker A: Sure, I try to have some hobbies. I've really loved this microsy adventure, so it does take up a lot of time. With that said, I've been lucky to live in a lot of awesome places. I went to college out in Colorado, so got very into the snowboarding. Get a little bit of that here on the west coast. If you get up into Tahoe in the beach. I've got three little kids and love to get outside, spend time with them. We've got t ball on Sundays these days, so trying to figure know how to run to first base or catch a those kids between the kids and Microsoft. See, it keeps me pretty busy.
[00:21:54] Speaker B: Hey, thanks for joining us, Matt.
[00:21:56] Speaker A: Absolutely, Jim. Thanks for having me.
[00:21:59] Speaker B: Our sponsor for this episode is Earhart Automation Systems. Earhart builds and commissions turnkey solutions for their worldwide clients. With over 80 years of precision manufacturing. They understand the complex world of robotics, automated manufacturing, and project management, delivering world class custom automation on time and on budget. Contact one of their sales engineers to see what Earhart can build for you, and their email address is
[email protected] and Earhart's kind of hard to spell it. It's Ehrhardt, and I'd like to acknowledge a three the association for advancing Automation. They're the leading automation trade association for robotics, vision and imaging, motion control, and motors, and the industrial artificial intelligence technologies. Visit automate.org to learn more. And if you'd like to get in touch with us at the robot industry podcast, you can find me Jim Barretta on LinkedIn. We'll see you next time. Thanks for listening. Today's podcast was produced by customer Attraction Industrial marketing, and I'd like to recognize my nephew, Chris Gray for the music, Jeffrey Bremner for audio production, my business partner Janet, and our sponsors, Earhart Automation Systems.
