Hello and good day to you from episode 21 of our podcast series Project Breakaway. A metaphorical and literal time in the day when we here at Predator cycling take some time away from working in the back shop to come and share with our listeners what we're doing, how we're doing it, what it takes to do it, our ideas, our innovative success stories and even our missteps and failures. If you find yourself with an interest in bicycles, composite manufacturing, out of the box design, or even curiosities beyond, I encourage you to stick with us. settle in and learn a little. I'm Courtney B, co-owner and project manager of Predator cycling. I'm here with my partner Arm Goan, the other co-owner, CEO, lead designer and engineer, and virtual collaborator at Predator cycling. Arm, we are excited to extend our conversation from last week about Nvidia's Omniverse software, which we dubbed Welcome to the Omniverse part one. And today, we're going to welcome our guest Mike Gyer from Nvidia to talk more about Omniverse from the creator's perspective. So let's begin our interview with Omniverse part two. Good morning and welcome, Mike. Hey, good morning. It's great to be here. Yeah, it's awesome to have you here with us today. Um, it's funny, Courtney and I were talking about the first time that we met. Like way back in Santa Monica. And um, I don't think I well, I don't I don't think you realize, but you were the first person outside of the bike industry to come to our shop and actually like, look at what we were doing like this is kind of cool. Like give us like an outside validation. So you are a very, very special uh, guest and uh, I don't know, it's it's very special for cycling. Well, that that means a lot. I it's funny with that uh, that meeting. I think about it all the time. Because I I almost bailed. I think because I had the wrong address. And I was kind of standing around uh outside the wrong uh the wrong shop. And it was like 10 minutes and 15 minutes and I was like, well, maybe I got the wrong spot and I I almost bailed on the meeting. All together, uh, but I I still remember just being completely blown away by the sophistication of what you guys are doing, uh, your vision for where you wanted to take composites. And um, you know, just the the passion at Predator around building high performance bikes. And so, um, I'm really excited to be here to spec out my new ride today. So we should jump in and start talking about that. For sure, for sure. Yeah, I'll take the order now. Um, actually, I remember, I don't know if you, when came to Santa Monica by yourself or with Kurt from Ansis. I I just remember Arm came home and said, this this guy came from Autodesk. And he just walked around and he didn't say anything. He just looked and he was real quiet. And I I I thought maybe he just was like really let down by what he was seeing. And I remember like, uh, it was funny the comment you made was uh, I I still remember. It's like, so what did you not make in the shop? Because we were showing like, oh yeah, we made this and we made that. And it was funny. I just remember that anyways. Yeah. Yeah, you you developed your own vacuum assembly and and everything. Yeah, it was a little a little much. But yes, yes, we did. It's it's pretty neat to see uh how Predator's progressed. And I got to say this is a a great podcast. The um, you know, everything the lighting, the furniture. Is this all real mahogany? It is. It is. It's all uh, it's we are not currently in the Omniverse, but uh, maybe later. We'll do we'll do the next one there. That sounds good. Great. So, um, let's push on. So, um, you're now at Nvidia. And I noticed on your LinkedIn profile, which I think you actually, because I'm a LinkedIn profile stalker. Um, I think you actually updated it recently. So, um, I noticed that you define your headline as building the Omniverse at Nvidia. That's your headline. That apparently is all you're doing at the moment. It's all I can think about. And uh, I got to say there's a a big team building it. So it's obviously not just me, but I'm uh, really excited to have uh, recently shifted my focus to to be much more uh targeted on the Omniverse stuff. And I got to say what uh the conversation you guys had here last week was was fantastic. Um, Courtney, Courtney, you really nailed it with the description. That's the only way I can understand things. Uh, you know, all the Marvel comic uh connection. It's uh, yeah, it's it's fantastic. Okay, so I'm going to regurgitate what we said on last week's episode just to um, explain what the Omniverse was again for anyone who might not have listened to that one. Um, so straight from the Nvidia website, they describe it broadly as a multi-GPU real-time simulation. collaboration platform for 3D production based on collaboration of Pixar Universal scene description and Nvidia's RTX technology. So it's a conglomeration of different design apps all interconnected into a virtual 3D ecosystem. Um and last week's episode Arm and I described our best definition of the software and its potential capabilities. Uh broadly across a lot of industries and also more narrowly and how we here at Predator, um can use it. So I just want to jump in and find out what your definition and Nvidia's definition of Omniverse is. Yeah, sure. Glad glad to. And uh I got to say, uh I agree with Arm. The way you described it last time is is actually perfect. Um, you're putting it in your own words. Uh, I guess to maybe just kind of give my own language to it. Um, the thing that we're hearing people be excited about. is the fact that that Omniverse uh gives them a way to connect the existing tools they use. Uh the different software applications. Um with uh some of the advanced AI and and rendering capabilities that Nvidia's developed over the years. And um allow anyone anywhere to portal into that. Um at the same time allowing people to write their own stuff on top of it. So it's kind of like taking the software world and turning it inside out. And rather than us handing you a box and saying, hey, go do these five things with it. Uh, we're we're giving you some foundational elements and and saying, hey. Go figure out what you want to do. And um that's where just in the few conversations uh Arm and I have had. It's been really exciting to see how uh his brain and uh, you know, the advanced manufacturing capability uh have taken the idea and and already um, you know, come up with things that that nobody else has. Uh, so we really see it just as a as a platform for kind of jumping off from. Yeah, every morning he tells me a new idea. What he can do with this software. I don't know if you want to come over for a sleepover so you can actually take it in. Because it goes way over my head. That sounds like a plan. Oh my gosh, that is that is true. 100% true. Yeah, it is. Um. No, it's it's cool and it's actually you guys have done a really good job of setting it up. Because like having it with like the way kit works and like create and knowing that like some of the tools that you guys have already built are built on top of like uh the kit functionality, like that makes it super. I don't know, for someone like me, I can look at it and go like, okay, like we could do something with this. Like we can build on top of it and add on to it and do what we want to do with it. It's cool. Um. Well, let's jump into the the five key parts discussed on your website for Omniverse, which is nucleus, connect, kit. Simulation and RTX render because you alluded to it. And we need to explain that a little bit more. For sure. Um. Yeah, I mean, the the USD side of it is, I mean. is the file format. And that's using uh what Pixar developed. That's open source. Um and everything is basically integrated into it. So like all of your your the the geometry as well as the the um the surfacing and textures and all that stuff is built into the USD. So it makes it, I believe easier for for um the the data side of uh Omniverse to work. Um so you can actually render it and and see the visualization of the parts. Um. The nucleus is a file server. Well, let let the professional here explain it a little bit better. No, this this is great. I think you're uh you're right on Arm. Um, you know, the the way we talk about nucleus. Is it's the the server, right? It's kind of the transparent part. It's the file store, uh allows you to share and collaborate. Um it can be installed on a a single laptop or in a a data center. Even uh, you know, running containers on the cloud. Um and then connect, this is is how we open the portals to other software tools. Um kit is is kind of your toolbox, right? For for writing your own apps, uh Python scripting. Uh really where your imagination can run wild. Um simulation are kind of the the um power tools or the components that you can add in. Right, all the AI algorithms for doing um inferencing and training robotics. And self-driving stuff. And then the um RTX render is is how you can. Pull in these massive scenes, right? I mean, you think about what's on screen with some of these scenes. Maybe like in Coco. From Pixar. Uh Coco from Pixar. You've got hundreds, if not thousands of artists working together on millions of different assets. Hundreds of thousands of different assets from different tools. And it's all got to come together. It's got to be perfectly orchestrated. Um I think that's that's why USD creates such a great framework to. to build all this stuff around. Um, but it's uh, you know, that that's the technology and that's that's great. What's really interesting though is is um, the things that folks like arm come up with, right? So I'm actually interested arm, what's, you know, what are some of the latest brainstorms you've had in the morning? Um, well, I a lot of the stuff that we've kind of been thinking on here on implementing it is is uh, or at least the biggest advantage right off the gun is is actually just the the the the data integration, just to be able to put all the data somewhere and store it and visualize it. Um, so like just sharing, you know, assets, um, across things. Like so like being able to take in stuff like, um, I think we're talking about it before, but about some of our machine data. Um, so we have all of our machines are designed and built. Um, we build them ourselves. And everything is data connected to the cloud. Um, so we have we have the ability to pull in machine data from our machines with all the models. So if we could stick that into Omniverse, um, and have assets everywhere, like spaced out and then updateable from our machine data, um, we can start building a virtual setup of our shop. And then start running, you know, plugging that entire assembly back into like uh uh a physics solver. So like maybe into like something like Ansis or we can plug that back into something else. Or workflows. Um, or using that for like maybe even for some of our product development side. Um, and start visualizing stuff better. Um, so I think it's like for us it's like, Courtney and I have talked about it in previous episodes quite a bit. Um, how simulation and um, our our digitization of our manufacturing system, um, has saved us so much money over the years. And we've able to actually like put out new product because of it. And Omniverse seems like the next frontier of building a, you know, digital twin. Um, so, I mean, that's kind of where we're, I don't know. This is where our head's going with it. Is Omniverse basically the digital twin concept? Is that broadly what the what was the birth of Omniverse? When you sat down, you're like, what is Omniverse? What are we going to do with it? Yeah, that you know, the back story is is pretty interesting on this thing. Um, and it it started uh around, you know, solving some of the the same challenges that that um, Predator is talking about, right, with a a concept of a micro factory. And allowing uh multiple people to teleport in, uh and bringing customers closer to the design and engineering process. So there was a Nvidia project called Holodeck. that was intended to be a a virtual or uh environment to do design, engineering, um, and and see the entire process through to the factory. And then close the loop back with data. And um, in developing that, we ran into some some pretty serious hurdles, uh, just with the tools that were available. And how they were architected to kind of narrow compute capacity down to a single game console or a a single computer. Uh, whereas, you know, we have um, the resources to bring multiple uh GPUs to bear against these things. And so, um, it's really something we built for ourselves. To to answer the question, Courtney. And we built it to help bring our teams together. To help us be better at testing the software that runs on top of our GPUs. Um, to help us uh better connect how we uh develop hardware. Right? Because we're a manufacturer as well. Right? At the end of the day, we we manufacture uh semiconductors. And um, and then also to connect all the the great stuff that's gone on in autonomous vehicles. So we kind of started seeing we've got these these core pillars of technology. And we're like, we should connect these for ourselves because it solves our own problems. Uh and then realized that um, you know, the market uh could benefit from the stuff as well. So, um, Yeah, I saw the demo at GTC. They had the uh the key uh the keynote. It was pretty cool. When they were talking about, is it uh, it's it's called drive, right? Or it's the the add-on for Omniverse for the autonomous vehicles. That was insane. You guys rebuilt the entire like your facility. Like the entry to your. It's amazing. Yeah. It was crazy to watch it. Um, yeah, that was super cool. Um, that's what inspired me to like, oh, like you guys did that. It's like, you know, doing that. I mean, obviously not for an autonomous vehicle, but like for automation of a factory. It's like that makes, I mean, it makes it look easy. So it's like, okay, like this is totally doable. Um, Yeah. I mean, one of the things for me that super exciting about uh working with Predator and and doing this omniverse stuff is just how you know, you're essentially you're doing micro factory, right? And you're doing highly customized uh all in house uh manufacturing. And you look at the challenges companies are having with global supply chains and I mean almost every week there's some kind of different supply chain disruption in in one business or another. And uh you know, what what Predator is doing to me is is the way it's it's going to go, right? Why have one mega factory in one country producing things for people all over the world, putting them on shelves and then hoping that they sell. Uh why not just make the thing that people need once uh where they live? And and do it that way. So, for sure. Well, and that's actually, so yes, and we're actually just having this discussion uh this morning about uh distribution. Like if we ever had to have the the Arm was talking, I was listening. If he says discussion, that's what that means. I was having a monologue. I don't buy it for a second. I don't buy it for a second. Uh yeah, we were talking about how um if we ever had to expand to like a fulfillment, because we're talking about fulfillments and and lead times. It was like, oh, we could set up a, you know, fulfillment network and have something on the West Coast to fulfill orders faster. And it was like, well, what's the point of actually having a fulfillment center? We would just set up another factory. Like, Mhm. It doesn't it wouldn't make any sense. Like we would just make the product there and then ship it. Um and it would actually be easier than trying to orchestrate shipping everything there and building it and, you know, putting it together. It's another benefit of of using composites, right? Is your your raw material and uh what you need to build stuff from. The the other thing that's interesting is is where you guys are going with additive and um you know, producing end user uh production parts with 3D printing and generative. And um that's the place, you know, where from the Omniverse and Nvidia perspective, we need companies like Predator to push us forward to think about how can this platform connect those kind of pipelines, right? How can we connect a distributed additive network to uh a network of uh engineers and designers? Yep. No, for sure. And that's one of the things that we've been playing with the idea and again, that's where Omniverse I think is going to play a big factor for us is um having that connection with our customers. Um and making a a better um being able to customize things digitally with customers um early on in the process, so we can actually work with them in hand in hand. So like, you know, looking at things on how we could have our entire asset library of our assets, of our models, designs, variations, accessible to us. We can build them quickly. Using things like Cloud XR with like augmented reality, um and being able to like display things to customers quickly. Um, and having them experience the product. And then, you know, we could either, I mean, for us, we have one place that would actually produce it and send it. But if you if you didn't, we could have, you know, if we had multiple places, micro factories, you could easily just disperse it anywhere and send it to the customer globally. Um, I don't know. That's kind of where we've been. That's where the thought process is going. We need more robots. That's the answer. I mean, it's it's pretty exciting just to think about how how much closer people are will be able to participate in the design and engineering of of what they use and make, right? I mean, you you see it uh you know, Courtney with your background in in film, you're starting to see some more participatory stuff in video games and Minecraft, people creating worlds. Um, you know, and to some extent you go on Amazon and you uh are, you know, picking options for the product that you order. Uh but there's no reason that that can't be extended a little bit to uh take more customization into account. And even to let, you know, non-engineers participate in engineering of it, right? At the end of the day, it's a a slider system of uh how you want something to perform and if you can incorporate um, you know, if you think about a bicycle fit, there's all sorts of stuff that that could be tailored there, right? There is. There's a huge that that's that's a a massive uh black hole right there. But yeah, um we've I know we've had a couple conversations about how uh fit can play a factor in in um in Omniverse and taking in real-time data from fit and uh and rider input to help optimize position, which in turn talks about sizing, which in turn, you know, depending on how you move, talks about your composite layup and structures and how things are made. And What's the thing that gets interesting is like the more data you have and the more you can represent the data, the better you can simulate things. Start using, you know, more AI, more, you know, ML systems. So that you can make better products. And that that loop is. I mean, we we were talking about in uh I think in the um um simulation worlds. When we were talking at our our a talk that we did together. Uh one of the things is like closing the loop on on data. So like you're designing something using AI. Um, you know, uh for a water bottle cage. And then testing it and then taking that test data and putting it back into the software. And trying to figure out where the problem is. You know, you start doing that now in Omniverse. Um, where you're not actually even going to make the product yet. Or you're you're simulating the process of making the product and then getting the results from that and bringing it back into simulation that you're doing locally. And then put that on top of actually making products that are shipped out and then pulling that data in. I mean, you're just getting so much data coming in. That you can actually do so much with. I mean, with AI and ML systems, you could you could go crazy. Well, and you you you what what it does is it lets the problem solving be handled by the AI. And the the human element is more about defining the problem space and and scoping what problem I want to solve. And then aim aim the AI and the ML at at going and doing the grunt work to figure these things out. So, so Arm and Courtney, how long till Predator produces the first fully AI designed custom bicycle? I don't know. Fully AI? I don't I don't think Arm would let that happen. He needs some say in everything. Uh, I don't know. It's. I think it I think the idea of AI for normal. Let's there's there's Arm people and there's Nvidia people. And then there's normal people. And I think AI sometimes scares people. So I think that there is a big learning curve to create education. For potential customers about what it can do for them and how it benefits them. And Arm is really good if you when we get calls of people wanting to make bicycles. And they have this old school mentality of like tube to tube construction. I just need this angle, this angle. And then he'll literally talk to someone for two hours on the phone. Um and describe all the tech that goes into the bicycle. And they'll listen like very intently. But then they'll be like, I don't know what you just said. But I want that bicycle. Yep. Um, so I. I think that there needs to be um a lot of more accessibility and education to this type of technology for just the regular consumer. Across the board. Yeah, that's a good point. But yes, an AI bike would be great. I mean, we're. So I'm trying to like you say completely done with AI. He's designing this in his head right now. But I think that's an interesting but I think that's like actually like a really interesting thing. Because like if you look at like our I mean, for instance, the bottle cage that we did. Um, I would say that that is entire I would say it's 90% done with AI. Um, the bottle cage that we did. Um, there are user inputs that we had to put in in the beginning to define everything. And I had my own bias that I I put in at the beginning and skewed it. Um, because I had settings that I wanted to do. Um. But it's close. And and the RF 20 is is also pretty close. Um. We haven't really gone into it much, but the way we do our layups on it is pretty automated. Um, from a layup structure perspective. But it would be really interesting to do what kind of what you were hinting at before. Is if you had a a builder. Maybe this is an Omniverse plugin for our site or something down the road. But like we had a builder on our site where you could do geometry. Setups, positions. Um, maybe even some sort of a little jet and scanner set up to scan a body to get dimensions off of it. Um, and then. Yeah. Put that in as to give a suggestion on size. And then a slider box, maybe questionnaire that does um riding characteristics and then based on weight and body proportions. We could design a layup schedule that was automated based on what your your inputs are. Wow. You could tune the fiber orientation. Yeah. We actually, so we. Anyways, one of the things that's clever about how the RF 20 is made. It's a mandril based. It's it's made from mandrils. So we can actually very accurately position fiber orientation. Um. And we use some bidirectional braided sleeves so that we can basically define angle orientation. By the circumference of the of the cross section. So by making some very small changes, we can very accurately position the fiber orientation in the bike. So, yeah, you could make totally a slider that literally goes there and says, I want X amount of deflection. I want this distance of fiber in order for the vibration to hit the frame. Which would be, you know, just how soft or how hard you want the bike to ride. Uh, yeah, you could, you, you could totally do it. And then automate out the pattern. And that's where I think like Omniverse could play a huge factor. Because you could have all of that patterning being made and then pulling that into that data. From, you know, like a, you know, fusion or something. That's pulling in that flattening pattern. And then, um, that's bringing that out to our machine that's actually plot cutting and cutting it. And then set up a little, uh, little five axis arm to pick and place to do orientations into kits. And then make the actual frame. And then the data from the frame. Um, that's being cured from our machine would actually upload everything into Omniverse. And data log all of our temperatures and cures. And validate those numbers against the simulation data that we ran in Ansis. To make sure the post processor is cured. And then the final simulation result of the load test that we did. Um. So yeah, you could. I mean that's. It just, it starts closing that loop. And then you could just think about changing the variations and things. And then figure out what happens. You know? Like what happens if I change the AC temp to, you know, 76 degrees in the shop? Like. How does that affect everything? That's pretty cool. That's pretty cool. So let's say I'm a, I'm an Olympic cyclist. And I'm going to spend 12 months training for the Olympics. Over which time I'll get stronger and my endurance will improve. Would you be able to have a bike outfitted with sensors? So that I could, I could maybe change the frame that I'm riding. Six times during the training. Or however many. Maybe twice even. I mean, we, we don't have one right this second. But so yeah, you, you could. And so, but the. I, I, I've kind of, so then talking about this. Because the more data we get, like the more we want more data. Um, but yes, so you could do it that way. But also maybe another thing and this is kind of where I've been playing with Omniverse a lot. Is what if instead of just taking rider data from like stress, like strain gauges that were embedded into the frame? Which you could totally do. That's not outrageous. Yeah. Um, but what if you did it in like standardized testing from like power transfer and power efficiency? Um. From like a power meter that's on your crank and then basing that off of your. Um, your speed, descending, accelerometer that you have on your bike built in and giving yourself an efficiency score. And then. Putting that into a a course that's already built into Omniverse and bench and benchmarking it against what should happen, what did happen. You can make a little avatar and like throw it in that swift game. Yeah, basically yourself and your bike and all of your. And stuff. And then bench line that. And then that kind of falls back into the idea of what we were talking earlier about like kind of fit. And then if you had a setup where you could actually fit a person on the bike. Or not even fit, but like, uh, uh, test. So you had, um. You know, some cameras so we can actually get positional accuracy and basically 3D scan the person. And find the deflection and while we're getting real time power. And then lay that against each other and bench line against different markers. Um, and then understand how fit plays a factor in fitness and the bike. Because it's, it's all, everything is connected. So like, you know. Was it the bike that did it? Was it the person that did it? Was it the training that they did? Was it the conditions in the, in the weather? Was it the humidity? Like what was it? And if you can kind of like create these bench markers. Then maybe we could make it and then you could make something smarter. And I mean. And then back to Courtney's idea of like what she's just talking about AI. I mean, I think, I mean, that's all using AI. You'd have to be using AI to build all of this. Um, but you still need the person to take all that data and kind of figure out what it's telling you. And how to build that out into the next stage. So. Anyways. That'd be pretty interesting. But. And maybe you could, you could get sensors in the handlebars. To determine glucose levels and blood oxygen and. Absolutely, you could. You could do that. You could also do like, uh, hand pressure, like comfort. To define comfort. Um. Yeah, there's a lot of interesting things. Um, you could do. I don't know. It'd be pretty fun. It'd be fun to play with. Yeah. So Courtney with the the film training in the background. Where does how does that intersect with the path for Predator? Oh gosh, I don't know. I'm, I'm all bicycles now. Um, but, uh. Yeah, I don't know. We're trying to come up with some virtual stuff and how we can get customers to log on and see more product. And. And we're not sure, we got to figure that out with our Omniverse. make them part of the story, right? Yeah, a way to get customers in here without getting them in here. Yeah. The podcast seems like a great way to do that. It's It's really a lot of fun. Yeah. It's been doing. The podcast has actually done really well. Um, it's trying we we we it's something that we we lost when we moved from Santa Monica shop where we had a lot of customers coming in and out. Um, and a lot of racers and a lot of stuff we used to do. Um, and as we started getting more and more into the manufacturing and some of our design stuff. Uh, world that kind of started going away. Um, and so we're we've been talking a lot about how to bring that back, that perspective of it of the of the business back. Um, yeah, so I don't know. We're still playing with it. But But Courtney, she she produces the entire company. She's the whole she's the producer for the company. So, That's amazing. Um, family manager. Family manager. So, but we should probably wrap her up here. Is there anything else that you want to throw in about Omniverse? Anyone? I do have a question. Who came up with the name Omniverse and how many comic books did they own? You know, um, I'm I'm pretty sure it was our our CEO Jensen. And I will just comment, I've got to say, uh, I think you guys touched on it last time. But the naming for this stuff is just great. Isn't it? It is. Nucleus and Kit and Isaac and Jarvis. Um, you know, as a as a nerd at heart, I just uh, I really love it. Well, it is. You know, it's great that it's um, you know, it speaks to what it is. But at the same time, it's not anchored in trying to play some SEO optimization game. Or uh, you know, try to, you know, it's it's part taco, it's part hamburger. Sort of things. It's like, use some imagination with it. So I don't know. I just I love it. I think it's it's uh, cool to see the naming be so uh, so relevant to where this thing is going. Yeah. Well, we can't wait to see what RM does with it here in the next year or so. We'll see. Looking forward to it. Great. Well, thank you for coming on and joining us. Yeah. Thanks for having me. This is awesome. Yeah. Yeah. Great. So, um, quickly before we wrap up, things to mention for Predator are bite cleat wedges. That we discussed previously, which is our second go-to market 3D printed product. They are available on our website, but now they are also listed on sale on amazon.com. And are currently on their way to warehouses for fulfillment. We thank you for choosing to take some time with us. And we look forward to future breakaways. Look for us on Instagram and LinkedIn, Facebook, Twitter, and in person here in Tennessee. We ask our listeners to please share, like, and subscribe. We're available on all major streaming platforms. Thanks for listening, have a good one and find some time to break away.
EpisodeMay 13, 2021 · 33:15
← Podcast
Project Breakaway with Predator Cycling
21: Welcome to the Omniverse: Part 2 Featuring NVIDIA's Mike Geyer, Ep. 21
In the second part of their "Welcome to the Omniverse" series, Predator Cycling hosts Courtney B and Arm Goan welcome NVIDIA's Mike Geyer to discuss the Omniverse software from a creator's perspective. Mike explains how this multi-GPU real-time simulation and collaboration platform connects various 3D design tools, leveraging advanced AI and rendering capabilities. They delve into Omniverse's foundational elements, including Pixar's Universal Scene Description (USD), and its potential for innovation.
Loading player…