“We’re on a mission to build climate resilient homes for all…leveraging the latest robotics and automation technologies.”

Construction costs have doubled in the last decade, and we have a shortage of skilled workers to build them anyway. In addition, climate events are getting more severe, so how can we start designing products that are climate-friendly, more resilient to changing weather patterns, and significantly less expensive? That’s Reframe Systems in a nutshell, explained Vikas Enti, CEO and Co-Founder of Reframe Systems. In this episode of the Manufacturing Matters podcast, Enti joined TECH B2B Marketing’s Jimmy Carroll to discuss the company’s mission, enabling technologies, and process. The discussion covered robotics, physical AI, 3D imaging, construction trends, and more. Check out this episode to learn how Reframe is leveraging industrial automation technologies to build sustainable homes.

Jimmy Carroll: [00:00:06] Hi everybody. My name is Jimmy Carroll. I’m the vice president of operations at Tech B2B Marketing. And welcome to this episode of the “Manufacturing Matters” podcast, where we discuss the latest trends and technologies reshaping the manufacturing industry. Today, I have the pleasure of being joined by Vikas Enti, the CEO and founder of Reframe Systems. And we’re live at Reframe Systems. First, thank you so much for having us here. Really appreciate it. It’s great to be here.

 

Vikas Enti: [00:00:30] It’s wonderful to to host you guys here. Thank you so much for making it out here.

 

Jimmy Carroll: [00:00:34] Of course. So for those who don’t know, please tell us a little bit about Reframe.

 

Vikas Enti: [00:00:38] Yeah. So, Reframe Systems. We’re on a mission to build climate resilient homes for all. Everyone’s experienced this today, where homes have become extremely out of reach. The cost of construction has doubled in the last decade and, unfortunately, we don’t have the skilled trades available to build them anymore. So combined with the labor shortage, we also see a challenge here where our climate events are getting significantly more severe. So how do we start designing products that are not only low carbon and how they use energy and how they’re produced, but also more resilient to the changing weather patterns? And how do we do that in a way where we’re significantly driving the cost of construction down, driving down the time it takes to build these structures, but also how do you make it super predictable? And that’s Reframe in a nutshell.

 

Jimmy Carroll: [00:01:25] I’ve got a lot of follow up questions that I’m going to get into, but I first want to ask: you used to be the former head of product at Amazon Robotics, right?

 

Vikas Enti: [00:01:32] That’s right.

 

Jimmy Carroll: [00:01:33] This is a shift, obviously. So what what inspired you? What led you to want to do this?

 

Vikas Enti: [00:01:40] The short version is I became a dad of beautiful twin girls. I think I was starting to subscribe to this life philosophy on how do you start living your life so that when your kids are teenagers, they look back and they’re proud of what you’ve done? And that was a whole mental shift. And I think that that gave me the clarity I needed to say, okay, what are the things I could be doing now? Like, I spent the last decade up to that point making same day shipping a reality for all, deployed over half a million robots across Amazon’s network. We did a phenomenal job in giving people time back to actually have the efficiency. And then the question became, okay, I’ve done a lot of work in giving time back. Are there things I could be doing to potentially change the mass equation? Right? We’ve been emitting carbon significantly more than we should be. Climate change was top of mind for a while. That point kind of catalyzed it even more, where I knew my daughters are going to be 19 and 29 and 2040 and 2050. What is the world going to look like at that point? Are there things I could be doing with the skill sets I’d built up over time to make a change? That search took about a year. And Reframe kind of came out of that.

 

Jimmy Carroll: [00:02:42] I love that. Like there are so many websites, so many companies today that you go to their website and you read their “about us,” and they have this sort of philanthropic idea of “we do this to better humanity.” But you’re actually trying to do that. I really love that. All right. I’ve got a number of questions that I want to ask. So, from start to finish… Right? How does it start? If somebody orders, somebody buys a house from you? What’s that process look like from start to finish? And how long does that take?

 

Vikas Enti: [00:03:07] It’s a really good question, and I think it helps also contrast what the experience is today. If you’re trying to buy a house, you’re either buying a home that’s already pre-built by a home builder or, if you’re a developer, you’re actually working with a GC or an architect to specify what types of products or what types of homes or buildings could be built on this parcel of land. How do you maximize entitlement and then drive value? Reframe’s taking a highly productized approach on how we think about homes and buildings as products. So today we have a duplex product type, which could also be a single family for certain lots. Then we’ve truly taken the New England triple decker, modernized it, and we have a triple decker product. And then in the future, we’re also going to have these small garden style multifamily buildings. So the experience here—— our customers today are developers. So we’re working with developers to make sure that they can go from an empty piece of land to a building showing up as quickly as possible. And the overall process is also very productized where they’re coming to us with the notion of, “I want to get a triple decker on this site.” And our goal is to be able to go from contract signing to permit submission in less than 30 days. Permits are not in our control today, but once you have permits issued, we’re able to go from site work and foundation built to an occupiable structure in as little as 30 days.

 

Vikas Enti: [00:04:23] The very first project we did, up in Arlington, was a small two-story single-family home. It was zoned as an ADU. We actually went from demolishing an existing garage that was on site to having an occupiable structure in 48 days. And that was the very first time we did it. We see pathways to go under 30 days, highly repeatably. And obviously it has significant benefits to the developer from a financing standpoint, use of cash, etc. But there’s also a huge benefit to neighbors and the whole construction process. Like our whole process, where we built up to 85% of the work, is actually happening in our factory. The factory we’re in right now. We’re taking all the noise and the dust and all the truck traffic away from the job site. So we’re actually able to be really good neighbors. The noisiest thing—— as an example, for our first project, we thought the crane was going to be the noisy thing. It was actually the food truck we brought in. The generator of the food truck was the noisiest thing on that job site. So we believe we’re also able to change the nature of construction and make it clean and quiet.

 

Jimmy Carroll: [00:05:20] Yeah, that’s a noise worth having, I suppose, right? I want to ask a little bit about the technologies. We’re standing right in front of a fairly heavy duty robot, and a really nice 3D camera and a lot of other underlying components that make it all work. Right? AI is a topic that gets talked about quite a bit, obviously in the general public and in the technology space. So I’ve got a few questions for you on AI. One of them is this idea of physical AI. So I’ve seen this term “physical AI” thrown out a little bit more and I have a sense of… You know, Nvidia, I don’t know if they coined it or not, but I know what they mean by it. And I have an idea in my head of what it means. But what does it mean to you? And how can physical AI be applied to building these homes?

 

Vikas Enti: [00:06:05] Absolutely. I think “physical AI” has been a really good way to think about the next evolution of “how do you go from robots that are learning but still require a lot of human input to design those systems, to a world where we’re actually able to have software that’s self-learning, self-creating to impact change on the physical world?” But the prerequisite for all of these things is the quality of data that goes in. So, in our world, like all the other cutting-edge robotics companies today, we’re using really advanced algorithms on our vision guided system. So we can use number two, grid lumber. We can buy from Home Depot and still build some really high performance structures. But there’s no magical algorithm we can use today that’s going to make things go faster. We actually have to start building the data set that’s actually relevant to our industry. And where we see promise is the pace at which algorithms are evolving——especially in how we think about motion planning for a robot that’s right behind me to how you would actually think about the overall process of even designing a home and actually having it run checks against permitting or zoning, be able to automatically create permit sets and construction documents. The thing is, there’s a whole set of tasks that take place in designing a home and manufacturing a home that we see significant opportunity for advanced algorithms to be part of it.

 

Vikas Enti: [00:07:21] They may or may not take the flavor of generative AI or transformers, but what we’re finding is that the amount of data that’s available and the way you actually catalog it, is making it useful for algorithms to catch up. Where we are in our stage today as a business, we believe that we’ve set ourselves up for providing the right data and context for these algorithms to be successful today. Even the way we use our robots, we’re actually not pre-programming the paths. Like our robots are actually learning the motion profiles they have to execute. So if it’s them picking an eight-foot section of two by four lumber, or if they would put up a 16-foot section of lumber, they’re actually able to recompute their motion plans on the fly to do that, which is very different than how robotic arms have been used traditionally. And this is the world, we would call it a software defined manufacturing, where we’re taking all the traditional limitations that come with automation, where you’re actually dependent on a singulated feed of material, you’re actually dependent on light sensors triggering a whole bunch of signal inputs for your automation… We’ve moved into a world where we’re all using vision systems to kind of get away from singulated processing to bulk processing of material, allowing the systems to actually make decisions on the fly. We’ve already made that step change. I think this is the world that a lot of us are living in today.

 

Vikas Enti: [00:08:30] Where this gets significantly better over time is now I could have different models of robots that are not as similar to the one that’s standing behind me, but they’re all able to share the same model set. So as an example, I might have a miniature version of this robot that’s on a mobile platform that’s actually able to go and do drywall, sanding, and painting, but it already has all the context and all the models that we’ve trained it on. We’re getting a fixed robot working. So we see a world where not only do we get to a scenario where harder tasks require more dexterity and more reasoning are getting easier to be performed, but we also get to a path where the engineering effort needed to build and design these systems are also going to be significantly shorter time cycles. We actually get to build these systems a lot faster than we could before. And that’s what keeps us super excited about this future. But I do want to highlight that it all starts with the quality of the data that you have, of course. And today we’re really mindful to make sure that we’re kind of setting the right infrastructure to capture all the data that’s relevant on the designing of the buildings, then translating that data into manufacturing instructions, all the way to exactly how the robot’s doing its task out here.

 

Jimmy Carroll: [00:09:34] So I have several follow up questions. One of them that I want to ask, though, is how——if at all——does does simulation come into play? Digital twins? When it comes to… well let me stop there.

 

Vikas Enti: [00:09:51] I mean, it’s a prerequisite. So today for us at a macro business level, the whole act of trying to build a home in a factory requires you to get to a high level of modeling, where the home is modeled to a level where every single fastener and fixture, light switches, all are modeled in it. We’re modeling the exact route, the wires, and the products your plumbing has to take. So in a way, we’re building the most accurate digital twin of the end product, which is atypical for construction. We have to do it as a prerequisite to make our entire process work. But then if you go a step back and say, “what do we have to do to make our underlying automation systems work at the factory level?” Today we have digital twins existing at the work-style level. So even this robot we took possession of in, I believe it was August 2023, by that point we had about five months of time——before we took possession of the robot——where we actually ran full simulation models of this robot to make sure we got all the motion profiles figured out, how do we design the full range of motion that’s needed to get this design behind it? We got a lot of work done in simulation to a high level of fidelity so that by the time we took physical possession of the robot, we could go from robot getting powered on to robot actually executing its first set of tasks in less than three weeks.

 

Vikas Enti: [00:11:10] And the robot that’s behind the camera here took us about three months to install. For folks who can’t see it, it’s this same Yaskawa robot behind me, but on a linear rail that can build wall sections of 16 feet or longer. We were able to commission that in less than three months, to go from hardware install to full production, because we could run through all of these iterations in our simulation platform and all the things that exist. And how do you add a seven degree of freedom for mobility, which is a whole layer of complexity that we’ve added to the system. But we could actually de-risk all of that in simulation. And I think we’re finally getting to a world where you can actually get true digital twins, where you’re actually able to keep your physical environment in sync with the digital environment. And we’re pretty excited to see what that looks like as the platforms become more robust.

 

Jimmy Carroll: [00:11:55] Yeah, so it feels to me——and correct me if I’m wrong, you’d be the expert here——but it feels to me in the last, I don’t know, 4 or 5 years, I’d say that there’s kind of been an explosion. Maybe not an explosion, but let me back up… 7 or 8 years ago, maybe even a little bit longer, AI kind of burst onto the scene in the industrial space, specifically deep learning. And it was being overhyped. I talk about this often, but it’s interesting because I always like to get other people’s perspective on it. At first it was being touted as this magic tool. “It’s going to solve everything.” “It’s going to replace traditional discrete rules-based algorithms” and things like that. But it seems as though in the last few years it’s really settled in and found really a number of novel and interesting ways to either be an enabling or complementary technology to existing automation. And in some part, it feels like that’s because of some significant advances that have been made in things like industrial computing and GPUs and simulation environments. Would you agree with that?

 

Vikas Enti: [00:12:52] Absolutely. I think a bunch of things have happened, I’d say circa 2016 onwards. I think deep learning as an aspect for training algorithms, especially on vision systems, has allowed us to do some really insanely good things really, really quickly, where now we’re actually able to allow arms to do vision-based pick significantly faster than the way we could do that with computational solutions before. And I think that’s been a huge advantage. What’s also come along really far in the same time period is the way all the systems that have come along for us to be able to capture images, catalog it, label them so that we can actually retrain algorithms quickly. These processes just become bread and butter today, that every college graduate coming out is learning these skills very quickly. But for engineers back in 2015, 2016, they’re having to translate from traditional automation techniques to using these algorithms, it was a huge step change. We got to see those live even at Amazon where we went from… We were trying to really realize the future of robotic picking. There was the Amazon picking challenge where it would take teams, right, like several minutes to make a pick. And that would be for like a very small subset of items. And five years since then, I think we’ve gotten to a stage where we can pick millions of different items in seconds. And a lot of that has had to do with the fact that the actual algorithms and the whole deep learning data sets have been available and are really pushing us to the forefront.

 

Jimmy Carroll: [00:14:15] Cool. That’s an interesting answer because I want to ask a little bit more about AI, but maybe in a different context. Two things: one is I have to ask this, like when you’re meeting somebody——I don’t know how old your daughters are, but if you’re meeting their parents or somebody from the first time, that’s sort of outside of your industry——and you explain to them that you deal in robotics and AI. Do you ever see any fear? And how do you dispel the notion that people should be afraid of AI? Because I think it’s the opposite, right? Like what you’re trying to do here specifically is trying to help humanity and not destroy it, like the movies tell you.

 

Vikas Enti: [00:14:53] Yeah. I think my whole mental model of robotics, software, learning algorithms, AI… These are all tools in our toolkit to solve really hard problems. And it’s understandable, given what Hollywood’s done, for people to be really, really scared about all the bad things that could happen if the tool was misused. But that is true for even a stapler that you have, right? You could always misuse a stapler and do bad things with it. But I think as it comes to people’s perception, I think there’s also, I mean there’s a certain level of risk and threats that are coming up when you think about the military use of robotics and drones, and especially with threat actors that are not necessarily friendly to the US, as an example. Like those threats are real. But if you kind of think about it from just an industry standpoint, like, should you be concerned about having a robot in your home? I’ve had a Roomba in my home for the longest time. It’s been a cute little equivalent of a pet, if not for anything else. Yeah, I think we’re finally going to see a future where we’re going to see more of these technologies available for general industrial use, general home use.

 

Vikas Enti: [00:16:02] I think it’s incumbent on us as companies using these tools to design solutions that are safe, well understood, and well communicated. And a lot of that is intent. We’ve spent a lot of time thinking about… Like in our case, today at least, we’re working within an industrial setting where our robots don’t have to work alongside people. Right? So at least the intent clarification is very straightforward, but we’re going to have to enter a new future where we’re going to have a lot more collaborative robots, a lot more interplay between robotic systems or just sensor-based AI systems that are working alongside people. I think there’s going to be a huge portion of user experience design that has to come in there for how do you actually let people understand the intent, so they’re not surprised if an odd behavior takes place there. But I’m an optimist by nature. My general view of the world is, we all have the right intents here. We have to start designing the right mechanisms to make sure that unsafe things cannot happen. And in terms of the intent of use, I think it’s incumbent upon us as a society to make sure we’re furthering and incentivizing the right intent versus the harmful intent.

 

Jimmy Carroll: [00:17:07] That’s a really good way of looking at it. I’ve talked to a lot of people, and I’ve had this ongoing conversation for years now with my in-laws, and they ask me all these questions about “what is it exactly that you do?” and “tell me more about the industry” and “tell me why I shouldn’t be afraid of robots” and “explain to me how robots aren’t actually taking jobs.” They’re creating them. And then we talk about things like AI and, in fact, I told my father-in-law yesterday that I was coming here and I said, you know, Reframe is a company that designs robotic microfactories that build sustainable homes. Hopefully I nailed that short description.

 

Vikas Enti: [00:17:40] Amazing.

 

Jimmy Carroll: [00:17:41] Thank you. Yeah. And he said, “Well, why are they doing that? They’re taking away the jobs of people.” And I said, “Well, not really, but I’m going to ask about that tomorrow.” So I guess I’d like to ask you about that.

 

Vikas Enti: [00:18:02] Yeah. I mean, like I said, the robots are a tool for us. All of these are tools for us to solve the problem for how do we make sustainable housing highly attainable? And the reality of the situation today is, even the reason we need a factory, is to have a more efficient process that allows us to increase labor productivity. Construction in the US today has about a half a million worker shortage. And these are all the skilled trades. And the compounding effect to that is about 40% of the trades are aging up. They’re set to retire in the next 10 years. And we’ve already seen the impact of cost and the pace at which we can build over the last decade because of the lack of workers and the lack of project cost making sense, and the fact that we don’t have starter homes anymore. A lot of it has to do with that. You can’t do brand new construction for less than $200,000 for a home. So when you’re actually undertaking a project, you just have to build $1 million homes instead to make the cost pencil, right? So the problem here is, how do you make construction more repeatable, fast, and cost efficient? We believe the biggest lever available to do that today is to improve labor productivity and to improve labor productivity we’ve taken two approaches. We have a lot of software we’re working on which allows us to give apprentices superpowers.

 

Vikas Enti: [00:19:15] So if you walk around our factory, you’ll actually see we have a bunch of high school students who are doing a co-op here, all the way to new carpenter apprentices who are all working off of iPads, which tells them exactly the information they need. Their instructions look more like Lego assembly instructions versus construction shop drawings. In fact they are 32–70% faster in performing their tasks. So we’re employing people at a much higher wage rate, and we’re able to make them perform more, so that we can actually produce more homes and solve one piece of the problem. The other piece of the pie is that there’s a whole bunch of tasks in construction that were a necessity, that are not really enjoyable. So when we pulled our carpenters earlier… Like to them, a lot of the pride they took on the job was like, “let’s work on things that we actually get to see,” right? So the finishing, the craftsmanship they can show off, the finished trim, cabinetry, etc. All this other stuff that happens inside a wall? They’re like, “hey, it’s a very repetitive process. We’re fine if that goes away and if it magically gets taken care of.” And we said, that’s a great requirement for our automation team to ask, can we robotically frame a home and kind of take the dull, repetitive task of framing away, so we can actually use the people we can hire to focus on more of the finished tasks and finish trims? And that’s where we kind of feel that the robot is still a tool in our arsenal for how do we make sure we can drive the overall productivity of labor up in return, produce more homes with the same number of people so we can drive the effective cost of homes down.

 

Vikas Enti: [00:20:40] And I think this is a construct that is very hard for human nature to like. It doesn’t pass human intuition very quickly because it’s very tempting to think about everything being scarce. Right? Like, scarcity is what triumphs in our initial thinking versus abundance. And it’s human nature to always assume that everything is limited. So if you’re taking some portion, if you’re automating certain parts of roles, it’s first instinct to assume that you’re actually taking work away from an individual. The reality here, and this reality actually also makes economic sense, is that there is a better, higher value work for those individuals to be performing to the point that robots and automation in general——and this has been true for the last 20, 30 years, robots are taking away of all the tasks that are really hard for people to do, really dull and or really dangerous. And I think that algorithm is going to continue for a very, very long time. Let me pause there. I know you had follow up questions on that.

 

Jimmy Carroll: [00:21:37] Yeah. First of all, I told my father-in-law, not as eloquent and detailed as thwhat you said, but I said, “well, they’ve got builders and construction workers that work there alongside. It’s not just a factory of robots doing these things with, you know, a faceless company.” But the other thing I said was, and I have to admit that I got this from your from your website or maybe your LinkedIn or something, but it said, hey, “home building is a trade that hasn’t really innovated much in the last 100 years.” So is that part of your intent here to say, “hey, we’re not just going to try and build something. If you build it, they will come. We’re trying to solve a specific problem, make things easier, more efficient and kind of take home building into the robotic age”?

 

Vikas Enti: [00:22:22] You’re right that home building and construction as an industry, while they’ve had significant innovation and adoption of innovation on the software side of things, or the way we design buildings, the way we actually look for interface checks with BIM (building information modeling), etc., there’s been significant progress there. But when it comes to the physical act of building stuff, that’s an area where power tools were the first huge step change that took place. Beyond that, I think we’ve had a lot of micro innovations that have taken place, but the industry as a whole hasn’t been able to adopt and make a significant improvement in productivity. A big piece of that is just how fragmented the industry is. If you think about automotive or almost every other industry that’s either building appliances or building chemical products or cars, the size of the company relative to the product actually makes a lot of sense, right? You could think of an automaker, you have maybe ten large automakers in the world that are building most of the cars. They own most of the market. That allows them to capture enough capital and resources to actually then put put resources against R&D. If you think about home building and construction. There’s 50,000 home builders in the US. The top 10 actually build almost, I forget the exact stat, like close to 40% of all homes. But then you have small teams, like five-person home builders that are then fragmented across the industry. So no single entity actually has enough resources to consult and say, we’re going to come up with a better way to actually build it. And unfortunately, that’s become the way things have gotten done over time, where the best construction companies are the ones that own the least amount of assets, that actually have the least amount of overhead, so they can spin up teams quickly to deliver projects and spin back down.

 

Vikas Enti: [00:24:01] So what this has done is built the most efficient project execution functions in construction, but not the best R&D functions, which require much longer time-horizon thinking, much different investment cycles. And I feel like that’s where it’s been a very interesting trade off for the industry to make. And this is where we see an opportunity for us as outsiders coming in. We have some of the naivete that comes in to say “we think we actually have a shot at doing this slightly differently.” And our approach can actually drive step-change outcomes. Like we’re not coming in to say, “we have to just fundamentally rethink everything end to end.” What we found is that a lot of the inputs are great. We actually believe the light-wood framing system is actually a really incredibly efficient system for how we use material. And we said, “let’s not change that. Let’s just figure out how to use existing materials, but more efficiently. How do we apply better techniques for how we use them, and then use software to make sure we’re taking waste out of every single step of the way?” So what we’re actually doing still here is back to what we did at Amazon. We’re taking time out. We’re taking time out of every single step of the process to go from selling a project, to designing a project, to manufacturing, to installing. And that’s where we drive efficiencies. And the only way to take time out is to bring technology in. And here we are.

 

Jimmy Carroll: [00:25:13] I like that, it’s really well put. I want to ask a little bit more about the sustainability factor here. And as somebody with your background and knowledge of automation technologies, what are some ways that you are either seeing or could see robots and AI and machine vision technologies, motion control, all the others——I don’t want to single out just those——leading to a more healthy, greener earth?

 

Vikas Enti: [00:25:39] Absolutely. The good news is we’re already seeing a foray in many different domains. Like, we’re trying to do this in home building. The reason we can build more sustainable homes is, one, we get the factory precision, which allows us to build homes that are significantly more airtight. But also the arbitrage we have here where we’re being more productive on the labor side, which allows me now to use higher quality materials that have lower embodied carbon, that allow us to still achieve similar cost parameters that make a project pencil, but with a much better carbon story behind it. So that’s been a place where automation is allowing us to, again, take time out, which is take cost out to use better materials. We’ve seen other industries——I think the most notable one, which we’ll all feel a benefit for, I would say is the recycling and transportation industry. We’re finally seeing the advent of robotic arms and vision systems that can sort, trash, or sort your mixed recyclables to make sure we actually have higher yield. And I think the more we can do that, like trying to have every individual take a practice where they’re making choices that can actually compound to become more efficient, is a really hard user-behavior to change. We’ve learned that time and again. But if you can use technology that allows us to be bolted on to end waste streams, as an example, where we don’t have to change user behavior necessarily, we should try. That’s a hard thing to change. But if the the cost and the time to actually sort through stuff in this case goes to zero, because we actually have automation techniques…

 

Vikas Enti: [00:27:04] And there’s one example where we can truly find a pathway to reuse material as much as we can. And I think the same applies if you kind of look at the stack up of all the other industries where carbon emissions are incredibly high, right? Like we gravitated to the built environment because our homes are emitting more carbon than our cars. And we found a clear path to apply our background and techniques in here, along with my co-founders. But if it wasn’t home building, I’d be in agriculture. That’s the other industry where the whole practices we have with agriculture today, with industrial farming, has really gotten to a scale where our soil isn’t able to actually absorb much more carbon. The amount of fertilizer use actually drives the second order cycle of carbon emissions and energy use that we believe, I believe, is actually… And anyone else out there, I’m happy to talk about this, but there’s a whole use case for regenerative agriculture where we can make the process of farming at an industrial scale significantly more efficient. But that requires us to rethink the entire process of what does the tractor need to look like to actually scale this up. And I think that’s, again, a place where not just industrial farms, but mobile robots, the whole gamut, the entire technology stack we’ve developed for manufacturing becomes very handy to have for the industry.

 

Jimmy Carroll: [00:28:12] Those are really, really good examples. And there are ones that I write a lot about, I talk a lot about, and hear and read a lot about. One is recycling. And, you know, hyperspectral imaging used to be a technology that was kind of complex, costly, large, kind of limited to the lab. But now you have cameras that are, you know, some of them are really *this big* that can be used to recycle just black plastics. And then in agriculture, I  like seeing a system like this or the convergence of different emerging or advancing technologies. I think agriculture is a really good example, because there’s been a lot of advancements in SWIR and multispectral imaging, but there’s also——like you’re saying——knowing where water, the crops… You can capture these images on smaller, less expensive multispectral or SWIR cameras. And then you can use AI to say, “all right, this part of the crop has been overwatered.” Or, “needs more seed” or whatever. I think it’s really cool to see applications where there’re a lot of different technologies coming together to solve problems. And you’re seeing these latest advances, like you’ve got a newer 3D camera in here, and then you’ve got a robot, but then you’re also using AI. And these are three of the technologies that are most interesting and talked about today. I’m glad to hear you say it.

 

Vikas Enti: [00:29:42] Absolutely. I think the world is bright.

 

Jimmy Carroll: [00:29:44] Yeah. On that note. What are some other technologies in the general space you’re in that you’re excited about? Anything that we haven’t talked about?

 

Vikas Enti: [00:29:56] I think we’ve covered a bunch of it. Like, I think to me, the biggest unlock, I’d say, is on the entire spectrum of vision sensors. So you mentioned different frequencies for different use cases. I think just the fact that the cost of imaging sensors have gone down significantly, and the deep learning algorithm allows you to actually get useful information out of it also allows increase in performance of a few orders of magnitude over the last couple of years. I think we’re finally getting to a world where… Like a lot of our work on making systems that were smart were really based on finding all these other workarounds to make sure we could replicate how a person or how a vision system would actually recognize the world. Everything from singulation, as an example, to all of manufacturing today, anytime there’s automation, you’re trying to singulate a component to a fixed location for a robot, to then pick it from that fixed location blindly and place it. That’s how we’ve done it for the last couple of decades, and that’s what’s resulted in massive manufacturing facilities. Just because the space utilization is awful. What gets us excited about vision systems are everything from the cameras on your phone, all the way to simple webcams that we have for surveilling the facility, they’re all getting to a quality of visual information capture that are useful for us to train algorithms to the point that I think we will truly get to this step change in the industry, where we’re moving away from singulated information processing and material processing to bulk processing, where all of a sudden all these sensors and conveyor systems and indexing systems that were the backbone of the industry sort of become less relevant and less and less necessary. Which I think will again translate to automation actually becoming more attainable for manufacturers.

 

Vikas Enti: [00:31:47] If you think about the old understanding, you’d say, “automation never makes sense for a small facility because you can never keep it utilized.” That made sense. If automation like a work cell like the one I’m standing in, costs you a few million dollars, you’d be like, great. Would never make sense in a small system, but we’d be able to scale the cost down to $150,000 for this work cell, where it’s equivalent to the rate of framing for two carpenters coming in. The reason we could do that is our vision systems actually allow us to get away from the need for singulation. All we can do is just roll in a cart full of lumber. The robot recognizes the studs, picks it up, frames it like that. Step change was impossible to think about five years ago, and now it’s just expected, right? I think we’re starting to see that progress. And to me, I think everything that’s built up of vision as a starting point gets us really excited here at Reframe. We believe that’s the backbone for most AI work that has to take place. Obviously, LLMs have taken all the attention lately. But I think where we start seeing the step-fold increase is like understanding vision and sensors have gotten attainable. And I think algorithms will also get significantly more attainable.

 

Jimmy Carroll: [00:32:52] I was going to close it out in a minute, but I have to ask another question just because you brought it up. So, yeah, automation is becoming more attainable. It’s getting easier to use. And ease of use is a term that gets thrown out a lot. Maybe that’s obvious, but things are becoming easier to use, to where a non-machine vision or non-engineer can operate a robot—— and even troubleshoot it oftentimes. Do you see Cobots as being a primary pathway for small- and medium-sized enterprises to be able to adopt automation?

 

Vikas Enti: [00:33:26] Absolutely. I think the number of opportunities for simple machine learning applications all the way through, I feel like there’s a whole bunch of applications that the industry hasn’t quite figured out a solution for. But if you start treating the robot as a toolset that a manufacturer can then use and figure out how they would use it, I think it starts changing the paradigm completely, right? And I think we’re seeing some really encouraging software systems, training systems, and the whole robot package coming together in a way that actually allows end users to figure out how they would use it. And I think that’s going to be a huge unlock for more solutions that an outsider wouldn’t be able to figure out, versus someone who’s living it day in, day out.

 

Jimmy Carroll: [00:34:07] Yeah, it seems to be the pathway for me. I just always like to ask people who are smarter than me if they agree with me. And I’m glad that you do.

 

Vikas Enti: [00:34:15] I mean, I’m smarter, but definitely agree with the observation. I think robots and I think more accessible automation techniques are the key for the US to reshore manufacturing here. There’s no other way for us to compete.

 

Jimmy Carroll: [00:34:31] Anything else that we haven’t talked about that you feel like you want to mention? And obviously we can let people know how to find more information. But if there’s any other general bit of information you want to put out there, please do.

 

Vikas Enti: [00:34:42] Yeah. Thank you. I’d say the mission we’re after at Reframe is dear to us. I think it’s a very personal mission. And we’re finding a lot of partners, be it customers, our employees, potential stakeholders coming in to join it. So we feel very fortunate there. We’d love the opportunity, selfishly… If you’re looking to build a home or buy a home, come talk to us. And the other thing I’d say, I actually thought you were going to be asking me about humanoids. That’s a question I get asked a lot, and I might as well address it. Now, in case someone’s thinking about what happens when humanoids become real and you’re actually able to use them. We welcome the day when we see different forms of automation that are more capable. The reality today is a humanoid, or at least most that are being advertised or on the market and have attention, can’t even lift a nail gun. They don’t have the capacity to do it. So I think we still have a long ways to go before they’re actually capable in industrial environments. Industrial arms aren’t going away anytime soon. Yeah, we should just view these as complementary tools on the tool stack.

 

Jimmy Carroll: [00:35:48] Yeah. And I’m glad that you did mention humanoids, I did mean to ask you about it. I caught a headline from the Nvidia GTC conference that Jensen Huang said that humanoids are closer than people think. I didn’t read the full article, but I’m curious to see in what capacity. Like, if it’s walking a trade show and handing out business cards? But, yeah, working in a factory? We’ll see.

 

Vikas Enti: [00:36:20] Yeah. I mean, there’s absolutely a bunch of, I’d say highly repetitive, low payload applications. Having come from the e-commerce world, there’s a whole bunch of tasks in there that you could see a humanoid with the right payload capabilities be able to handle. I still think there are more fundamental issues to be solved here in terms of like, what is the battery capacity on these robots? Like, are these going to be like the early drones where you had 10 minutes of flight time and you had to recharge stuff? I think there’s a whole bunch of engineering that has to be solved. There’s merit to solving these. We think there’s a future to it.

 

Jimmy Carroll: [00:36:57] But the other thing, too, is that I think about and I’ve heard talked about is, why humanoid and not just a wheeled robot with arms? That’s what I’m looking forward to learning more about and seeing. Obviously, I want to see these humanoids. I want them to succeed. But those are the robots that people are most scared about. So we’ll see how the general public reacts to that. But like you said, it might be a little bit far away. There’s not too many of them yet. Although… Did you see the recent video of Boston Dynamics’ Atlas robot moving around doing breakdancing moves? It’s pretty impressive.

 

Vikas Enti: [00:37:37] It is always fun to watch their videos. And I think it’s a matter of time before we can find useful applications for them and scenarios where you’re able to perform a set of applications, like… One thing that’s really important to distinguish here with the new class of robots that are coming up, a lot of the work we’ve been doing in manufacturing, industrial automation is using a robot to perform a specialized task in a specialized, controlled environment. But that’s where you start incurring all the second order costs that are involved in the amount of space it takes. Have you got cordoned off stuff? If you could have automation systems that can co-exist and work alongside people, but also perform a set of tasks versus a single task with the same hardware that’s on that robot or that automation system, that would be the superior solution over time. And I could see how one such bet is a humanoid or arms on a mobile platform gets us closer to the reality where you can perform a set of tasks versus a specialized task. And I think we should all be welcoming that, because it’s going to make industrialization a lot more efficient. But the reality check for today, though, I think, is there is still a lot of hardening that has to happen. The capabilities of these systems also have to increase to take on a broad range of industrial applications. So the capability might be getting there sooner than we expect. We’re taking a more cautionary note here, in that there’s still going to be a long tail of edge cases that have to be solved before you see a humanoid in every factory, every use case. But there are definitely use cases in our factory here where we welcome partnering with the manufacturer that has the right payload capacities on their systems for us to go take on a set of applications that our software can control and kind of move that forward.

 

Jimmy Carroll: [00:39:25] Yeah. I’m looking forward to it too. It’s very exciting. Before I close out, tell people what’s the URL to your website if they want to go learn more.

 

Vikas Enti: [00:39:35] Yeah. Check us out on Reframe.Systems or ReframeSystems.com, whatever your preference is. And there’s plenty of information there.

 

Jimmy Carroll: [00:39:44] And if anybody has any questions for Vikas, “Manufacturing Matters” would be happy to pass those along. Or if you’d like to send in any comments or join the podcast, please do. Manufacturing-matters.com. Otherwise, thanks for listening or watching.