If you’re a first-time climber trying to tackle Mount Everest, the only way you’ll make it to the top and come back alive is if you’ve got a great set of Sherpas to carry your stuff and show you the way. That’s how Ted Larson, CEO and co-founder of OLogic, views his company’s role as it helps startup companies navigate their way from lab to volume production. 

Building a good working prototype in an academic setting can be a far cry from building tens of thousands of production-grade products. But it’s not just startups that need help and guidance along the way. In this episode of Manufacturing Matters, Larson sits down with TECH B2B Marketing’s Aaron Hand and Dan McCarthy to share stories of opportunity and challenge from two decades in the trenches helping electronics and robotics manufacturers develop successful products. 

Aaron Hand: [00:00:01] Hello and welcome to this episode of “Manufacturing Matters,” where we talk about the technology and trends that are affecting manufacturing today. I’m Aaron Hand. I’m with Tech B2B Marketing, and I’m here with my colleague Dan McCarthy.

 

Dan McCarthy: [00:00:15] Hello.

 

Aaron Hand: [00:00:17] We’re joined today by Ted Larson, CEO and co-founder of OLogic. Welcome, Ted. Thank you for joining us today.

 

Ted Larson: [00:00:25] Hi. Thanks for inviting me. This is great.

 

Aaron Hand: [00:00:29] We like to keep the introductions very minimal so that we can hop right into the discussion. But if you would spend a couple of minutes just letting our viewers and listeners know what OLogic is and kind of the space that you play in, predominantly.

 

Ted Larson: [00:00:45] Right. So we’re a product design firm. We’ve been around for 20 years. It’s hard to believe we started it in 2005. I would say the three primary markets … We started out mostly in consumer electronics and toys. And then we moved heavy into big industrial robotics and mobile robots in the mid-2010s. And then in the last five, six years, we’ve done more like Internet of Things. But either way, it’s product design for those kinds of markets and helping people build successful products and get them shipped and out into the market.

 

Aaron Hand: [00:01:23] Okay, great. I think our plan today is we’re going to focus on the robotics primarily. OLogic has really had an array of customers, I think. For our followers who know that robotics space well, they’re certainly going to be familiar with companies like Fetch Robotics or Plus One Robotics. But you also have some real household names in your lineup. You’ve worked with companies like Google, Nvidia, Hasbro. So how do the challenges that startup clients bring to the table differ from some of those bigger, established players?

 

Ted Larson: [00:02:04] Right. So there’s kind of like two kinds of startups that we tend to encounter. We encounter ones that have built a product before, and they have a little bit of understanding of what is entailed in that. And then we have other ones that have absolutely no clue how to build a product. And there’s a big educational exercise that has to go along to kind of teach them everything that they don’t know. And in some cases, those kinds of customers can be really, really great, because they really accept mentorship and they really want to learn as they go. In others, it can be challenging, where maybe they’ve come out of a large organization. They worked at Apple, they worked somewhere else where they were part of the product development cycle of a product. And then they think they understand what the process is. But, you know, without the infrastructure of one of those big companies making it happen, the process is very, very different when you’re trying to do it on your own as a small startup. So, trying to educate them, saying, “oh, all of these things that you think that are going to happen automatically? They’re not. You’re going to have to do them yourself. And here’s how you’re going to have to go through it.” So that that generally tends to be the big tradeoff. One of our taglines that we throw around all the time is, “look, building and shipping a product is like climbing Mount Everest, especially if you’re a first-time climber. And the only way you’re getting to the top and coming back alive is if you’ve got a really killer group of Sherpas to carry all your stuff and show you the way.” And that’s what we feel like. We’re in that position frequently.

 

Aaron Hand: [00:03:52] Wow. Yeah. Okay. So, I’ve interviewed you —— just so viewers know —— I’ve interviewed you a couple times recently for a story I was working on and it’s just really interesting to hear some of your war stories from this space. You work quite a lot in robotics, and there just seems to be a lot that can go wrong there. So, especially moving from academic lab to commercial development, can you share with us some of your recurring horror stories from developers as they try to navigate that transition?

 

Ted Larson: [00:04:31] Yeah. So, in the case of an academic lab … I don’t like going out and airing other people’s dirty laundry, people’s names. But, yeah, you can go and look through our portfolio and see some of the projects that we’ve worked on. And you can go read some of the news stories that have happened as a result of some of those projects that … They sometimes don’t tend to end well when they start in academia. They maybe have a very high-profile, charismatic founder, who has no problem raising a lot of money. And, in academia, they’re used to, you know, building a few really good working prototypes and then publishing a paper and then that’s it. That’s all they’ve got to do. And that’s a very, very different animal than building a production-grade product, where you’re going to manufacture tens of thousands of them, perhaps, and then have them all delivered to the customer in good quality working order, and then the customers are happy with the result. Right? We usually try to encourage people to do a bit of beta testing and other things to get that market-focused feedback built into the product.

 

Ted Larson: [00:06:02] Academic projects usually don’t have any market-focused feedback cooked into them. Maybe they do, because they did some research with people out in the field, which is great. But, surviving contact with the customer and getting that customer review of what they thought it was before you’ve really given it to somebody who’s, like, got a more discerning eye as to what to review? Like a product reviewer or somebody like that? You’ll learn a ton of stuff that you had no idea about, like how they’re planning on using [the product]. A lot of customers on the consumer side, they’ll do stupid stuff, and you’ll be like, “oh, I never could have anticipated they would have tried to do it that way. That makes sense. We need to build in some process or some things into it to make it easier to use.” And, the other item we often find with, again, big charismatic founders —— they tend to make big promises to their investors and their customer base that possibly could never be delivered on at any point. We’re kind of of the mindset that’s, “ship a product, get it into the market, get it doing really well, then ship a follow on product.” In the case of consumer electronics? The consumer industry gets pretty bored. Within 18 months, you need to be shipping something new that is a revision of [your product] or an upgraded version of it or something that offers some new features or capabilities that the prior version didn’t have.

 

Ted Larson: [00:07:56] And that’s usually based on customer feedback you got on the first version that you shipped. And that also leads to better customer retention and things people are more willing to upgrade or buy the next version of, because the prior version was so cool and interesting. Now the newer version does something new. So, we’ve got a lot of customers that have had very successful success stories. Those are the ones that I tend to trot out the most and be like, “oh, here’s how they started. This is how they’re evolving and this is how they’re improving what they’re doing.” So an example of one of our really early customers from, I want to say it’s like almost 10 years ago now, we worked on a product from a company called Wonder Workshop. It was kind of like a children’s toy robot that was designed to teach software engineering to kids. And they started out kind of focused more on the home and on families.

 

Ted Larson: [00:09:13] But then they realized there was a huge opportunity in schools and curriculum and building curriculum around the robot and then adding small little add-ons to it to make the curriculum new and make it more relevant. And they’ve done tremendously well in just making these small additions, these small features that you can buy, new curriculum that goes with the robot. And they continue to sell more and more and more of them and schools and educational people tend to be, I think, the vast majority of their sales these days. So I look at that and I’m like, “they did it correctly.” And they’ve been able to sustain themselves over a long period of time. Versus other ones, they’ll build something really splashy and it’ll go really good. They’ll raise a big pile of money. But if they really didn’t build a market-focused product, then they’re trying to back into how to make it market focused. But it’s kind of too late at that point. And then they can’t sustain enough revenue in order to actually be a viable business. They have to continue to raise money —— and, you know, in this economy that we’re in, or whatnot, sometimes we go through these periods where it’s very hard to raise money and then they’re stuck.

 

Dan McCarthy: [00:10:34] To jump in for a moment. It’s eye opening for you to talk about planning two moves ahead, particularly after you’ve released the product. I want to go back to something you were talking about earlier. For some of these less experienced companies that might not have the playbook: It sounds like what you’re describing is methodologies for project management and for launch. Right?

 

Ted Larson: [00:10:54] Yeah. Right.

 

Dan McCarthy: [00:10:55] Testing, all the stuff that, you know, they don’t plan ahead or they don’t realize [they need] just to get the first product off the ground. Do you have a playbook? Does OLogic provide that sort of service? I mean, are there common mistakes that you can talk about?

 

Ted Larson: [00:11:10] We do. I think one of the things that’s common —— and it’s funny, we literally just had this conversation yesterday with a new customer. A new customer comes to us with a working prototype. They had the prototype made by some prototyping group in China, but they didn’t quite end up with what they were hoping for. They needed to evolve it, to turn it into something that they can produce and where it’s going to meet more of the needs of what they want to do. But there’s this issue of like, “okay, we’re taking the thing that they started with, we’re learning the lessons that they learned.” Because it doesn’t matter how you start one of these ventures, right? You go and spend money doing something to get things moving and build a prototype or try to make some handmade thing. There is tons of learning that has happened already. So the first step is to try to understand what learning has happened, what mistakes were made, how do we not repeat any of those mistakes, and how we can help them to evolve to a plan that we know will succeed? And then the second piece is to make sure that all the requirements you’ve sat down and you’ve thought through, these are all the requirements that this product needs to fulfill, based on my knowledge of this market out here, and I’m going to build a specification for how I’m actually going to achieve those requirements and what is going to be achieved now versus what are we going to do later. And it’ll all be based on cost and trade-off and customer need and how many people want it and whatnot.

 

Ted Larson: [00:12:57] That kind of analysis, to do that in the very, very beginning is incredibly valuable. And a lot of people skip all that with this “if we build it, they will come” method. And then maybe some people come, but then they can never get to the next step, right? Because they’ve built this thing that’s highly constrained and where they didn’t really match the requirements of what they constructed to what the market demanded or wanted. So, you got to do a little bit of market research. In our case, our playbook is: first step,  really good requirements documentation. Sit down. Think it through. A really good specification for how we’re going to meet those requirements with tech or engineering or whatever. What items are we going to address now? What items are we going to leave to later? And then we produce some kind of really crude MVP, that’s that kind of first go at it. And then put it in front of some kind of customer who would use it and get a sense of, “well, how do they interact with it? What happens? And what kinds of things are we missing?

 

Ted Larson: [00:14:13] But you’re doing it with that early, early thing. You’re thinking about it like a focus group. Think about putting it in front of people who aren’t your friends and family, because those people are going to lie. They’re going to tell you, “oh, it’s great.” You know, like, “everybody wants one of these.” Don’t take it to that. If you’re a college student and you’re doing it, go set up a table in the quad with the thing and say, “come try this thing and tell me what you think.” That kind of stuff. You’ll learn so much from that brief interaction with you, with your thing, and then talking them through about what it’s supposed to do and what value it adds. Then after you’ve done that, you bake all that back in, and then you build your actual first prototype. And that first prototype that you build —— in our case, we build it with this intent that that thing is going to get made in a factory somewhere down the road. It is not just the first prototype of a thing that may get thrown away and completely started over. So you’re kind of putting a stake in the ground that you’re going to follow a path to a factory to build. And that’s kind of our methodology and we’ve been very successful with that. We’ve shipped hundreds of products that way.

 

Ted Larson: [00:15:32] In some cases, too, we’re not doing the whole thing. Like maybe they did a bunch of work already, maybe they even have a product in the market, and they’re looking to do a 2.0 product, and they’re really unhappy with how they went about doing the 1.0 product. And they need fresh perspective on  how to do this better this time, so they’ll have a better outcome. We have a couple of customers where they’re already shipping products in other markets and they’re wanting to build a new one, but they’re coming to us to help them do it. So, yeah, we definitely have a lot of process that we bring to the puzzle because we’ve got 20 years of experience doing it, and then we share that process heavily with our customers on [things] like, what are the steps to get to where you want to go? And what are the holes that we identify that you’re missing?

 

Aaron Hand: [00:16:34] I did want to get back to what you were saying about introducing some new, cool things to the market as you go along —— and you do have those consumers, in particular, who want to get that new phone every year or two.  But I would think it’s a considerably different story if you’re selling a robot to a manufacturer. For example, do you have that same kind of iteration cycle?

 

Ted Larson: [00:17:02] Yeah. So, this is another interesting thing, when you think about how robot companies’ business models work in comparison to consumer business models. Right? Consumer business models usually are: you build a product, you make a big pile of them, you sell them to individual consumers, they purchase them. Maybe there’s some kind of online component. The hot thing these days is there’s some kind of an online component to drive engagement with it. And then maybe there’s even a subscription for a premium version of it. So, even my sous vide cooker in my kitchen? You know, I buy it, but then I can sign up for their online thing, and I can get access to a much better database of recipes or things to make it do things. Or I can choose to use the free version, which, you know, maybe that’s all I want, right? So, there’re a lot of people wanting to build this data and analytics platform piece behind the gizmo. I would say that a lot of consumer products are moving towards that, which allows them to get AI and things into their product, where maybe they wouldn’t have been able to in the past.

 

Ted Larson: [00:18:22] And then on the robot side, that business model tends to be very, very different because these tend to be low volume, high complexity items. The manufacturing of them is tough. And the question is, how do you deal with the maintenance of it? In most cases, they don’t just sell the robot. There are companies that sell a robot, but most of them don’t JUST sell it. They offer it as a service. And so, essentially, they’re renting you the robot or leasing it to you, and then you have some kind of a maintenance plan. But included with that —— like, next year, let’s say that they do a bunch of upgrades. You can get the upgrades included in your system, and you’re not having to buy a whole brand new one, because it’s an expensive piece of equipment. And so there are certain companies that have done very, very well with this “robot as a service” model. An example of one that we worked on is the Bear robot, right? Like, if you get the Bear food delivery robot, you’re a restauranteur, it’s more of a service-based model, so you don’t buy it.

 

Dan McCarthy: [00:19:36] You’re essentially buying it as a service. And then they put it in the restaurant, they get it all set up for you, and then they help deal with any issues that come up. And if it breaks, they send a service person out to fix it. It’s all just part of a service plan. So they do it that way, and there’re a few other big ones that do it that way. Also, investors these days really love software-type models. They love seeing recurring revenue. Selling individual end-product is tough to have recurring revenue unless you actually have some kind of back-end service, some kind of a cloud service. So it’s all about, “oh, what’s your ARR (annual recurring revenue) and what are you planning for that?” And, “how’s that going to work?” So most of the business models that are being pursued these days have some ARR component to it. But the ARR piece is usually a software play or it’s some full-system maintenance play, where they’re getting ongoing maintenance from the thing that they’ve delivered.

 

Aaron Hand: [00:20:53] Okay.

 

Ted Larson: [00:20:54] But, I think about it —— again, from a manufacturing standpoint —— these tend to be products that are low-volume, high-complexity. And they’re making them here in the US, too —— that’s another item as well.

 

Dan McCarthy: [00:21:15] So to your point about these being low-volume, high-complexity robot systems or systems —— not just all the component technologies, but the system design, how they come together. The component technologies are evolving pretty quickly, from vision, AI, all the networking to controllers … Put your designer hat back on: Are there particular design challenges when all those component technologies are evolving so quickly from when you start the project to when you end the project? Is there a way to future-proof that? So when it goes to market it’s not immediately obsolete.

 

Ted Larson: [00:21:56] Yeah. So, software is a big one. You know, the ability to have an over-the-air update mechanism and a mechanism that allows you to put new software in the system and unlock new capabilities. I would say there are certain traditional robot industries that are going to get worried soon, or they’re starting to get worried, because …  The way that you would deploy robotics in a factory setting, for example, the way that that has been traditionally done for decades is, you buy a robot from a manufacturer, you buy it through an integrator, usually. The integrator comes in, they set it up to do the task that you want it to do in the factory, and then they leave it alone. And then it just does that task all day, every day. And then if it breaks, you call the integrator, they come back and they fix the setup. But it’s basically an amalgamation of a bunch of off-the-shelf parts that were essentially screwed together to solve that specific manufacturing task. So, let’s say that you now want it to do something different than it was doing before. Suddenly they’ve got to come back in and reprogram the whole thing to do this other task. And that could take weeks, months maybe, depending on if there’s any additional mechanical fixturing that has to happen.

 

Ted Larson: [00:23:44] And those have been popular in things like car factories or, you know, certain manufacturing lines for pick-and-pack for years. Right? And they’ll continue to be relevant there. I have a friend that runs a company that builds robotic palletizers, right? So all they do is take boxes off of a conveyor belt, arrange them on pallets, and then wrap them and then get them ready to go on a truck, right? And he goes out to the Midwest to these companies that are doing packaging, and he’s just gleeful at how woefully low-tech they are. Right. And you would think, okay, we’re in Silicon Valley or whatever, but the Midwest is incredibly low-tech in these packaging companies and stuff. We see this kind of stuff here, but go out to Iowa or Nebraska or whatever and go to a company, go to a soap factory or some place and —— especially if it’s not an ultra mega-producer, it’s a small producer —— they’re not very high tech. They have people packing the boxes and stuff. They don’t have robots doing it.

 

Ted Larson: [00:25:17] And it’s also evident if you ever go to … There’s a great robot packaging trade show called Pack Expo, where all you see is all of the robot packaging technologies that are out in the market. It’s a fantastic little show to go to. But, when you go to a robot arm manufacturer … Okay. The typical method for how it does a task, it’s all about programming repeatability. So the arm gets it here, it moves it here, it moves it into some other fixture, some other thing. It does that over and over and over again. Most of these systems have no vision-guided feedback or anything. It’s just, go from here to here and do it repeatedly, over and over and over and over, thousands and thousands of times with millimeter grade accuracy. And the way that they achieve that, at least in the automotive sector, is big heavy metal robots that you can’t get anywhere near because they kill you if they run into you. And that mechanical accuracy has come through mass. Okay. So now we have, you know, AI people. And what they’re doing is they are taking the world’s crappiest robot arm and making it do things that the big, heavy, gnarly ones do.

 

Ted Larson: [00:26:50] They don’t do it as deterministically as the big, heavy, gnarly ones do. But many tasks don’t need highly precise repeatability. So if it’s just, putting a bottle of soap in a blister pack, okay, that doesn’t need to be millimeter accurate. You just lay the soap in the blister pack and you’re good to go, right? But you need to know that you’ve actually put it in there, right? So I’ve had a kind of a funny joke that I’ve done where, last couple of times I’ve gone to a robot trade show in [the past] year or so, I’d be like, “I don’t want to buy a big, heavy, expensive robot arm. I want to buy the crappiest, cheapest, least accurate arm you’ve got. Which one is that?” Right? And of course, none of them are like, “oh, no, we don’t have any of those.” And it turns out they’re all going to have to have one. And it’s coming. And there are some companies that are starting to produce these things. There are people that are coming to us, asking us all the time to make them ones from the ground up, because they want to be able to do it really, really inexpensively —— because, again, a typical UR5 arm could run you 35,000 bucks, right? A cheap, crappy one that, maybe it’s reliable and it doesn’t break, but it’s made out of really inexpensive components. It doesn’t weigh a lot. It’s not terribly accurate, but can be serving, pretty reliably, the same location over and over again? You know, that could be made for $1,000 and that’s a different animal.

 

Ted Larson: [00:28:32] And the robot arm guys have completely missed the boat. Like they have no … They’re still going after that high-accuracy, big metal sale. So I would say the last big trend in robot arms was collaborative robotics, ones that collaborate with you and they don’t kill you when they bump into you. The NEXT big one is really inexpensive, inaccurate, collaborative robots that can then be used with AI and vision systems. And I’m still kind of waiting for those to appear in the market. I think it is only a matter of time for them to show up. People ask me to design them for them and we do some of that work. But, I haven’t yet seen a major manufacturer, like a KUKA or a Universal making what I call the ultra-cheap, crummy robot arm that doesn’t break easily but isn’t very accurate.

 

Dan McCarthy: [00:29:38] That’s a good tip. So for all our investor listeners, the future is retrograde.

 

Ted Larson: [00:29:43] It is, it is! The future is retrograde. Going back in time.

 

Dan McCarthy: [00:29:51] So, you said a lot of these don’t require vision, which is ironic because you kind of undermine my next question, which is to zero in on vision a little bit and talk about whether the vision factor, how that factors into the work that you’re doing for robot developers. Are there challenges there today? I mean, does vision also become unnecessary with the….

 

Ted Larson: [00:30:13] No, no. So, vision is actually “the way” that you integrate AI and robotics, right? It’s the glue that holds AI and robotics together. They have these things that are called VLMs or VLAs, vision language action models, the key being “vision” and then “action” being robot arm movement, or robot movement. So, there’s a lot of really amazing work being done for how to train systems using vision, to train AIs, using vision to make a task repeatable and then build a foundation model that then can be added on to do a new task. The idea is, you build an AI foundation model for a whole host of manipulation tasks. You then use that model in an embodiment with an arm that may be different than the one that they used for the original training of the model. And then you do additional training to just add the new tasks on, with not a lot of repetitions —— maybe 10 or 20 or 30 repetitions —— and then suddenly now it can do that task.

 

Ted Larson: [00:31:42] My favorite people that I watch very, very, very closely are … There’s this company up in San Francisco called Physical Intelligence. And they are making a foundation model for vision language, action-type robotics. And they have a phenomenal video of a robot where … They have a robot that they’ve built for what they call “doing tidying,” right? So it’s not like cleaning, but it’s tidying. And they have a video where they brought a robot into four different Airbnbs that it had never seen before. So, it had never been in these spaces, and they said, “tidy up the kitchen.” And it just went and found the cabinets and put the dishes away and cleaned up the stuff in the sink. And it did it really well in all of the demos that they gave. Another one was like, “make the bed” and it grabbed the sheets and changed the bed all around and picked up all the stuff on the floor and put it in the hamper. So, it’s tidying. It’s not full on cleaning. But, I think the most important part about it was, they were dropping the robot into an environment that it had never seen before. So that’s really amazing. To me, that’s  groundbreaking.

 

Ted Larson: [00:33:05] One of the biggest problems of all? Is the training data problem of, like, how do they collect a lot of this training data? And there’s a thing … I’ve been working on a blog post. There’s a bunch of research that was just published out of Meta, of them using people wearing cameras and then doing manipulation tasks, and then using that to train robot data so that a robot can do it.

 

Dan McCarthy: [00:33:36] Cool.

 

Ted Larson: [00:33:37] And I think if they figure that out so there’s not any specialized gear in order to do it, suddenly this stuff’s going to accelerate even further. The humanoids and stuff are going to be able to do a lot more, as we have more of this kind of data collected. So, those are the areas that I follow pretty close and I get really excited about. And I wish more people would come ask us to build their mechanical embodiments. It’s mostly a software problem — you know, the AI is mostly a software issue. But if you’re working on one of these things and you need help that understands the AI piece, call me because I’m super enthused about the whole meshing of robotic hardware and AI, and how do they all go together, and how the piece is going to get glued together. Those things are really the items that are really hot on my hit parade every day.

 

Aaron Hand: [00:34:31] We’re at this point of, you know, “AI is great, but it can’t take my trash out.” So that robotic embodiment will let the AI take your trash out, right?

 

Ted Larson: [00:34:42] That’s right, that’s right. I remember, a number of years ago I lived in a house that had a really steep driveway that was up to the street. And every Wednesday, I had to go out and drag the bins up the hill to the street. And every time I thought I was going to just have a freaking heart attack dragging those bins. “They’re going to find my lifeless corpse by the garbage can,” you know? And, we still don’t even have that. I think, “when’s that coming?” So, yeah, I’m hopeful that something like that will come in the future.

 

Aaron Hand: [00:35:26] And that vision is a key enabler for your crappy robot arm, right?

 

Ted Larson: [00:35:31] That’s right. The crappy robot arm could do it if it was attached to the right mechanical thing, yeah.

 

Aaron Hand: [00:35:38] I think you need to start putting ads out saying, “hey, do you need us to build crap for you?”

 

Ted Larson: [00:35:44] Right! That’s right. “Let’s build the world’s crappiest robot arm!” You know, how crappy can it be? But people don’t want to do that. Like, they react funny when I say, “well, where’s the really cheap, crummy one?” I almost kind of hope that they’re going to pull it out of some box, some haggard old thing that’s like, you know, that they played with that didn’t go anywhere. But they don’t. None of them. They’re all built, they’re all based on, like, ultra precision. And then safety is a big item too, which is true. But safety’s also less of a problem if the payloads that you’re manipulating are small —— which is things like the dishes in the kitchen or whatever. I’m not moving around 100-pound boxes of screws. If it’s just simple task-oriented items, pick-and pack-tasks, the kind of arm that you need for that is not the kind that will kill you. So that’s why I’m also skeptical, too, when I see people with humanoid robots and they’re doing videos and humanoid robots are doing parkour or they’re doing handstands or somersaults or whatever, or dancing. Breakdancing. I don’t need a robot to breakdance. I need a robot that can do fine manipulation tasks, like pick up that wine glass and put it away without busting it. And it turns out most of them can’t do that. And that’s the reason why they’re showing them breakdancing. Because it looks cool. It’s great PR, but it’s not… It doesn’t really serve any kind of useful purpose.

 

Aaron Hand: [00:37:37] So when you give them a wine glass, you’re also going to see some breaking and popping, right?

 

Ted Larson: [00:37:44] Probably. But again … she’s not 12 anymore, but my daughter, when she was like 10 or 12, and I’d say, “hey, clean up the kitchen,” right? She’s going to break some things. It’s an understood … You know, it’s a risk. You know it’s not going to be perfect every time. Just like we people aren’t perfect every time. So, hold the same expectation out for the robot that you hold out for yourself. Again, unless you’re putting it in some kind of precision environment where, you know, the repeatability has to be perfect 100% of the time. That’s a different animal.

 

Aaron Hand: [00:38:19] We do joke about that, the really cheap, crappy robot. But, in fact, a lot of your development is very concerned about how much each of those components within the robot costs, right?

 

Ted Larson: [00:38:33] You bet, you bet.

 

Aaron Hand: [00:38:35] So you think about using those off-the-shelf components, correct?

 

Ted Larson: [00:38:39] You do. And, actually, it turns out that the way that you really reduce cost in a robotic product is, don’t ship a robotic product that is built up of a bunch of off-the-shelf components. Take the expensive pieces that are really core to your success or failure, and build bespoke versions of them that are specifically suited just to that task. We have certain vendors that we partner with that I wouldn’t recommend building a bespoke version of that to put in a robot project, because they do it very, very well. So, for 3D cameras, we’re partnered with ORRBEC, and we’re also partnered with Intel on the RealSense. Both those products work very well for 3D camera systems. Trying to cook your own from scratch is a mistake, because they work very well —— and for the price, why? Why try to cook it again? But, a motor controller? I cannot tell you how many bespoke motor controllers we’ve built, because… The type of motor they’re controlling and what they’re using it for and all that. There’s off-the-shelf ones that do support 100 different kinds of motors, but [our client] only needs to support one in the one product that they’re shipping. And the cost savings of a custom-made motor controller versus an off-the-shelf? It’s an order of magnitude.

 

Ted Larson: [00:40:12] So we do pretty good business in building custom motor controllers for robot products where they are planning on shipping volumes of them. You know, if you’re only going to ship 10 or 20 or something, great. Just make it out of off-the-shelf parts. If you think you’re going to make thousands, then you need to go look at the parts that you can make bespoke, specific bits of that will give you significant cost savings in any kind of scale. I would say another area that we provide a lot of assistance is, identifying what components that you have in an off-the-shelf prototype that could be replaced to go at scale. Another example: we’re partnered with Nvidia. Uh, we’re an NPN (Nvidia Partner Network) partner. Nvidia has a wonderful ecosystem of companies that take their latest Jetson SoMs —— you know, their carrier boards for their Jetson SoMs —- to use them as a brain in a robotic project. The problem is, is that most of those carrier boards provide access to every possible thing that the SoM could ever do on that board.

 

Ted Larson: [00:41:30] And maybe the customer’s only using a camera and a thing, and they don’t need all that extra stuff. So you could end up in a spot where the carrier board… In many cases, if you go out shopping, you’ll find the carrier boards for most of the Jetson SoMs are actually more expensive than the SoMs themselves. You know, they could cost you a couple thousand dollars for just the carrier board. And so that’s a place where we get a lot of business, or people come and say, “oh, we’re doing a robotic product. It’s got a Jetson SoM in it. We only need to use these features on the SoM. We don’t need all of these other connectors and all these other capabilities. Can you make us a custom carrier board that just gives us access to the things that matter?” And then the price of that carrier board goes way down. And so that suddenly makes putting Nvidia as your CPU brain inside of your product much more palatable and much more capable. And that’s one of the reasons why you see Nvidia’s doing a lot of things to show off, you know, “these are people that did do that and here’s the amazing results that they’re achieving.”

 

Aaron Hand: [00:42:46] Yeah. And that’s interesting, because my brain automatically thinks, “oh, well, you want to use off-the-shelf components to make it cheaper.” But that makes perfect sense in cases where you really need to create a bespoke product. Talking [00:43:02] about [00:43:02] Nvidia, I actually wanted to bring up: I was looking around on your website, and looking at some of the projects you’ve done, and saw the Kaya robot that was really focused on showing how hobbyist components could make the robotics more accessible. But from having talked to you before, it seemed like you were kind of pooh-poohing the whole “hobbyist components that might have come out of an academic lab” and needing those components to be more reliable or more performant or what have you. So was there something about Nvidia’s approach that was an exception to the rule, or were there predictable speed bumps that you had to deal with with those components?

 

Ted Larson: [00:43:47] So, really, more than anything is … Again, when I look at the big semiconductor companies that we work with all the time, like MediaTek or Nvidia or Intel, they need some kind of a setup where they can get some user adoption and where they can get the academics, the hobbyists, the home tinkerers… Where they can get them to even look at using their stuff for a possible solution. And so from a development perspective, development kits that support all kinds of interesting things that you could do with it are a fantastic starting point, right? Because it makes it more accessible. You could just get this kit, build this thing.

 

Ted Larson: [00:44:51] In the case of the Kaya robot, it was like, “oh, we have the Nvidia Jetson Nano, but what am I going to do with it? What’s an interesting project I could do with it?” And so in that case, Nvidia was like, “hey, could you help us make a robot that, you know, the whole design is open source. Its 3D printable?” You could go download all the parts, get the bill of materials off the web, and build one yourself in an afternoon. And, what’s at the heart of it? It needs a Jetson in order to work. So you go get the the Jetson Nano and it’s designed to fit in the thing. And now you can suddenly take your Jetson Nano and make it drive around and do interesting things. And I think, having that kind of developer ecosystem of interesting things that you can do … If you can’t even do that, how can you start to visualize the next steps of, “how does that lead me to building a product?” And I think what they want to do is, they want to make sure that they’re in the boat before it leaves the dock. And that is a way to do it. So, the same way that we are building developer boards for MediaTek … We have a variety of developer boards for building for them. How did we build them? We put them in a Raspberry Pi form factor, based on their Genio line of chipsets. But you’re basically able to buy other off-the-shelf Raspberry Pi form factor add-ons and plug them into those boards and they just work. So, again, it makes it more accessible to that developer ecosystem. And then when the person’s like, “okay, I want to make a real product about this.” Great! No problem. We can show you the path for how to go from there to there. It’s okay to start with hobby parts. It’s just not okay to try to ship a product with them.

 

Dan McCarthy: [00:46:55] You’ve spent enough time with us, it’s been… We could talk all afternoon, about this. This is a great topic with you. So I have one quick question and we’ll kind of wrap up, but we’re curious: we’re talking about you trying to balance those off-the shelf components to kind of control costs. Is that going to be become more challenging with supply chain uncertainty?

 

Ted Larson: [00:47:15] I would say, off-the-shelf parts does create a supply chain problem, because, again, you don’t really know how things are going to evolve. When people get worried about supply chain —— China supply chain is a common one that comes up. There are certain components that come from China that, it  doesn’t matter if a 400% tariff is placed on them, you’ll still be able to get them cheaper than you would be able to get them here. So you do need to spend some time looking at that. I would say, China supply chain is a thing where —— and this is in general … And this is another huge piece of advice with doing business in China: There are a lot of people that buy things from China. They put them in their products. They don’t really ever get to know the supplier in any way. They don’t really understand, like, where is it actually coming from? If you have components that are critical to your product success, and they’re coming from China, for sure get on a plane and go meet them. Go there. Build a relationship with them. From all my time in China and factories and other things —— you cannot build something in a Chinese factory without spending a bunch of time in China,  hanging out with them, and building that relationship with them. It’s a relationship based on trust.

 

Ted Larson: [00:49:01] And, you know, you’re going to trust them to do the right thing, and they’re going to call you when they run into a problem. And that trusting relationship is critical, I would say. And vice versa. They need to trust me that I’m going to pay my bills and that I’m going to give them the things that they need, and I’m going to work with their team to get it. I am baffled when I see entrepreneurs trying to stand up a Chinese factory on a product that they’re working on, and they’re trying to avoid going there, and it’s like, that’s a recipe for disaster. Just don’t ever do that. Like, that’s the one thing! And, then, same thing goes with supply chain items. You’ve got a motor manufacturer that you love, that you bought a few motors from them out of China. You think they’re great and you’re going to use them long term? For God’s sakes, don’t just not go there. Go there! See their process. See how they do it. Understand what their struggles and challenges are. So then that way, if there’s a problem that comes up in the future, you’ll be aware of it. Let’s say they’re having problems with the government over there or heavy metal magnet acquisition or something. They’ll tell you, and then you’ll be you’ll be ready for it if there’s going to be a problem.

 

Aaron Hand: [00:50:22] Yeah all right, that makes sense. Well, I agree with Dan, we could talk with you all day. It’s always very interesting. Ted.

 

Ted Larson: [00:50:30] Yeah. Would love to do this again sometime, this is awesome. I’ve enjoyed meeting you both very much.

 

Aaron Hand: [00:50:37] All right. You too.

 

Ted Larson: [00:50:39] It’s very cool.

 

Aaron Hand: [00:50:40] Thank you very much for your input and insight, Ted. And, thanks also to our viewers for joining us on “Manufacturing Matters.” If anyone has any follow up questions for Ted, please put them in the comments below, whether you’re on LinkedIn or YouTube or wherever you’re watching. And just a reminder that you can see past episodes of “Manufacturing Matters” podcasts at our website, which is manufacturing-matters.com or on your favorite podcast platform. So in the meantime, please hit favor, hit like, subscribe, and continue to tune in. Have a great day.

 

Dan McCarthy: [00:51:16] Thanks, Ted.

 

Ted Larson: [00:51:16] Thank you.