EP 319 – Dushyant Verma – CEO at SmartViz – Redefining Precision With AI: A New Age for SME Manufacturing

by | Apr 10, 2024

"We are in a very interesting time where we will see the embrace of technology in a much more rapid rate…We want to be the first in the industry to be delivering a truly autonomous machine vision platform for manufacturing." - Dushyant Verma The Asia Tech Podcast engaged with ⁠Dushyant Verma⁠, a co-Founder and the CEO at ⁠SmartViz⁠ to explore the intersection of AI and Manufacturing.

Some of the topics that Dushyant covered in detail included:

  • The integration of artificial intelligence (AI) into manufacturing processes represents a pivotal shift from manual inspections to automated, precise quality control.
  • The exploration of how technology empowers manufacturers to evolve from being process-focused to data-centric entities.
  • The often-underestimated significance of small and medium-sized enterprises (SMEs) in the manufacturing ecosystem.
  • The global nature of the opportunity for AI technology adoption in manufacturing.
  • The transformative impact of modular technology on small and medium-sized enterprises (SMEs).

Some other titles we considered for this episode, but ultimately rejected:

 
  1. Revolutionizing Quality Control: The Future of Smart Manufacturing
  2. AI and Modular Tech: Transforming the Manufacturing Landscape
  3. From Manual to Autonomous: The Evolution of Manufacturing Processes
  4. Redefining Precision: AI’s Role in Flawless Production
  5. The New Age of Manufacturing: Quality, Efficiency, and AI
Read the best-effort transcript

Read the best-effort transcript below (This technology is still not as good as they say it is…):

Michael Waitze 0:04
Okay, let’s do this thing. Hi, this is Michael Waitze. And welcome back to the Asia Tech Podcast. Dushyant Verma, the CEO of SmartViz is with us today. Dushyant. Thank you so much for coming on the show. How are you doing today?

Dushyant Verma 0:16
I’m doing well. And it’s a pleasure and quite an opportunity that you’re presenting to us tell you something about me about SmartViz and manufacturing.

Michael Waitze 0:24
cannot wait to learn more about it. Can you give our listeners a little bit of your background for some context? And then we’ll jump into the main part of this conversation? Yeah, oh, yeah.

Dushyant Verma 0:32
My background has always been in product tech. And data. I’ve worked a lot with customers worked a lot with my partners and distributed technology. And the SmartViz, which was more like a serendipity. To me, I was listing a number of these manufacturing facilities on a product that was handling, these were like small and medium factories had very good quality product. But again, when I’m putting my product into the hands of the customer, I need to be very, very sure about consistent quality, not just the one. So when you go and visit them the processes and everything is okay. But when you go back, what is the check? So as my dear friend who’s now I see you and Rob, what should we do? How can we control the quality in a consistent way? Right? He said, Why don’t you take a picture, because now what you would do is you can utilize the visual technology is so advanced, and you would know everything that was the start of our journey. And we realize that this problem of asset and its quality, the operational data is not just limited to the kind of factories that I was looking at. But it’s all cohesive from the last one year we’ve built a model. We’ve built a technology that can democratize the data understanding of these patterns, and therefore detect defects and the animals. This is always in there for a large manufacturer who can invest wait for technology, but technology is very nimble, very fast, and get that can be turned around in in this time. So this way, I think what we will be enabling this new age is to enable such manufacturers to adopt technology fast, utilize it to understand their processes, monitor their assets, and therefore provide much better quality than paper.

Michael Waitze 2:08
So what do we do when we use less expensive cameras at scale, we allow them to be given to SMEs. So smaller businesses, as opposed to large corporations that have kind of like an unending amount of money right to be able to do this. What is the innovation here? In other words, how are we now allowed to use these smaller cameras? What’s the hardware and the software kind of combination to make this innovation possible? A

Dushyant Verma 2:28
great question. Because most of if you look at the large corporations, they invest in heavy duty large size cameras, that enables them to take even smaller defects, right, but a small manufacturer or medium manufacturer, they are constrained not just for the cost, but also for the time, they do not have, say a quarter of a million to put down in terms of investment over nine plus months for its investment, what we’ve been able to do is we’ve been able to break this traditional machine vision into smaller modules that can be utilized at scale. And it gives them a faster way to adopt the technology, we can roll out a solution in about six weeks time. So that is fast. And that is because of the proprietary technology that we have, we just need about between one to 20 images. And we should be able to create a model help you identify the effects and also, more importantly the animal, right and put that back into wherever system that you want me to start or PLC, however you want to utilize it. And this is done in six weeks time. So now what our manufacturers have the ability to do is not really wait for those 689 months. But to experiment with this process, see if it works for them. Because we are able to implement it fast, you get a fast ROI, and see whether it works for you or not. So there is no a big investment that you have to do. So in terms of technology and our USP, I think what we’ve been able to do is to make it modular earlier, which was like a full scale full suite of products, we’ve been able to make it much more smaller, right. So if you really want to just measure, you can go ahead and measure or if you want to take more complex visual inspection use cases, such as getting color oxidation, or finding out some defect interest you have the ability to and also we give the ability to do one thing first. And then second thing later. So you can start with the measurement and then pay as you go along. This flexibility is what this customer needs. And that is what we’ve been able to provide to innovation.

Michael Waitze 4:23
Yes. So this modularity sounds like it’s one of the biggest innovations can you talk as well about I don’t think we can have any conversation anymore today about technology implementation of technology and software without talking about artificial intelligence, right? And if you’re saying you’re moving people from what was used to be like manually just look at the end of the production line and say this thing looks okay, which is not a great result, right? To actually employ technology to do this, this speed, some of the speed and some of the iterations must come from using AI like what role is it playing here and and how do you see it changing the way these types of inspections get done at scale?

Dushyant Verma 4:54
So there are a couple of questions in your question and they’re always stopped doing that but you So I would like to step back and let me give you the provides AI has always been there. I mean, we do it subliminally, whether we’re using Google, Facebook or say any of these platforms. But what chat GP has done is to make it much more ubiquitous and give the definitive intelligence. Now, what has happened is, and big tend to get these questions a lot from our customers. That is, can after some time, if I ask a question, would your system be able to provide that answer? The GBD is a language model, the result that it shows it does simplify the life but you do not know what’s the source of the data. Right? Now the difference between the AI that we are working on, you have the data, and that is your own data. And what you do is you slice and dice in different ways to make it much more meaningful, you will always have the efficiency data or quality acceptance data. But now what we tend to give with this is a lot more meaning to the data that you always have, and get it more consistent. We digitize this process, we make your asset more understandable, try to correlate it with the processes around so that when you find a defect, you just don’t know that it’s a defect. But to go further down, you know, and try to find out why the defect happened, right? We do it on the production line, which means you have the ability to stop and understand and then proceed and not really waste energy, manpower, and time and replicate those errors, just

Michael Waitze 6:21
so I can just jump in for a second. Sorry, the implication for manufacturers, it’s actually really large, right? Because you make this point about catch up to these large language models, they just go out and get data from anywhere, right. And, frankly, the results, we don’t know where it comes from, we should but if you have the manufacturing data yourself, so your data is segregated. The things that then it allows you to do right, if you’re building the proper data lake or the proper data management is not just looking at the defects, you’re right, but as to figure out where they came from, but also to create more consistency in the product. So even a small manufacturer now can keep up with what a large manufacturer is doing from a consistency and a quality perspective, like it almost changes the whole game. No, absolutely,

Dushyant Verma 6:57
you’re perfectly right in this and what has happened in the last few years. Because of COVID, large manufacturers have gone autonomous, they don’t want to be in a similar situation. So with that, without manpower, they can’t function. Now what’s happened is not this technology is percolating down. And the expectation is much more higher for now the manufacturers to deliver. So like for our immediate customers, they provide to larger manufacturers they provide to Tesla they provide to John Deere, Caterpillar. And when we go to them, they’re one point brief to us, guys, we don’t want to send out even a single defect, number one second, when these guys come, we just don’t want to show our processes, but also the underlying technology, that really good the reliability of our processes. We don’t want to be something inferior. Because we know that if we don’t do so, the next three, four years, we will be replaced by a factor which is one. Yeah. So one thing you would agree with me, Michael, that that technology is changing in such an unexpected way, I mean, manufacturing in us, you just have to keep pace with it. In fact, what has happened is even in Amazon warehouses, which we think is like a state of art, after three or four years, they really think back in terms of technology is really giving you the right kind of ROI at the right interesting thing. So we need to be very cognizant of this, we need to keep pace with this, we are able to provide something which is a lot more less time bound does really solve the problem to the core.

Michael Waitze 8:20
Do you want to comment at all I like to I like to make this point, actually, whenever we talk about SMEs, right, because I think people’s opinions, because it’s in the name small and medium sized enterprises, right or small businesses, the word small is there. And I think people think that the impact that they have on the local, regional and even global economy itself is small, but in Asia, in Southeast Asia, and I think this is true in most places, 90 something percent of all businesses are SMEs. And I think that that’s true in the manufacturing space, too. So again, just back to the implication of the tech is so large, because it allows a small manufacturer in Thailand, in Malaysia, in Vietnam, in the Philippines, to modernize the way they build and produce things way faster than they could have ever done this before. In a way that’s not that expensive. No,

Dushyant Verma 9:02
absolutely. You’ve just done the right point, Michael, and even I thought that SMEs are like, really, really small. But yeah, actually, you know, in our research, the SMEs contribute close to about 67 of the total manufacturers production, which means a manufacturer would be responsible only for say about 30 45%. The rise 65 comes Wow, small and medium. This is like an ecosystem without which the manufacturers can survive. I mean, we do inspection for small components for differentials for precision engineering. And all of these are critical components. I mean, we heavily provide to the automotive sector. And each of these companies, even a single error can learn them contract manufacturing to big problem. I mean, we’ve heard of so many recalls, one recent was the Honda I airbag, which itself could just be about 10 billion and not taking into account any penalties or blacklisting for this right. So this is huge. The implication of the impact of these small guys, I wouldn’t really want to According to what their primary medium pocket size is close about 100 million kind of a turnover for Singapore and about 30 million in India, they are large, but they still have the same inefficiencies as a small guy, they still want faster technology, they still want cost effective, fast ROI, kind of a solution at the same time, they are have the ambitions of the bigger parent manufacturer. So I think we are in a very interesting time where we will see the embrace of technology in a much more rapid rate. And so in the last few years,

Michael Waitze 10:32
yeah, it’s pretty exciting. Can I ask you this? Where are you based?

Dushyant Verma 10:35
I’m based in Singapore, Michael. And my team is based in India, how many people we got close about 15 people, three of them have hardware specialization. And the rest are data scientists and AI. And AI engineers

Michael Waitze 10:47
got to be an incredible team. Sorry, go ahead. Yeah. So I

Dushyant Verma 10:49
think we will be one of the few teams which have got a unique combination of both hardware and software. And because we see more and more that hardware is now coming to the forefront, unlike what it was about 10 years back where people were talking about software about SAS. But now even companies like Google start manufacturing phones and video, you can see the stock price. So it’s a great age where software and hardware have to comply to each other. And so therefore, we’ve tried to build this culture right from the start. There are hardware engineers and software engineers, they work in tandem to solve this

Michael Waitze 11:19
problem. Steve Jobs used to quote a guy and I can’t remember his name, I was just trying to look at what you were talking. But the guy who said something like if you’re really serious about building your own software, you’ve got to build your own hardware as well. And I think it’s really important. And you’re right, it’s really coming into the fore right now. And I think Nvidia is a really great example of this the NVIDIA story we can spend hours talking about it’s a different story. But it’s really interesting, too. Can you talk about where you think the biggest opportunities are? Is this a massive Indian opportunity? Is it a tie opportunity? Where do you see this stuff playing out?

Dushyant Verma 11:47
I would say this is more like global opportunity. Because after the China thing, a lot of manufacturing has shifted down to the Asia and India. So we’ve got partners in Vietnam, in Thailand, in Indonesia and Malaysia, where the extra production that used to go to China is now coming in huge respect for China, in terms of the technologies that have been applied these geographies still invite if I may estimate, we will be at least five to 10 years behind what they have. I also believe that now the technologies increase at a much more rapid pace with AI, and the kind of models that we can work with. So the opportunity is truly global. We also have a couple of instances and discussions going on in US where us is a great example, they are averse to employing people. And we are in talks with design mold designing outfit, they say that if I employ someone would cost me about 5000. And then there are obviously laws in terms of fighting it. Like an embedded system, if it is a menial, manual process. For us the immediate opportunity because of our startup because of our funding, we would still be more in Asia and India. But given the nature of the problem that we’re solving, I truly believe it’s a global opportunity. We have huge pockets of opportunity in both us as well as Europe.

Michael Waitze 13:03
It’s interesting though, because this idea of disaggregating or disambiguate a lot of the manufacturing outside of China as sophisticated and as amazing as the Chinese manufacturers are right their political reasons, economic reasons, geopolitical reasons, and stuff like that, for some of this some of this movement that’s taking place, but the US is actually after decades of not manufacturing anything not even like chewing gum. Now they’re starting to manufacture a lot of stuff there, again, for strategic reasons, but also for economic reasons. I see that as a gigantic opportunity, too. Can we talk a little bit about the status of smartphones, you mentioned funding, you say you have a team of 15 people? So have you been funded? How are things going how’s the business itself going as well,

Dushyant Verma 13:42
I think we’ve had a dream run in the last one year since we’ve started, we never thought that we could make it this week. So we ended to we’re still very small. We got incorporated in Singapore last year. In about five months of our operation, we delivered 43,000 USD revenues. What we want to do this year is about close to about a million. And we have a couple of good discussions that are underway with some of the big guys like Landis, Nissan, the two wheelers and electronic space, we’ve been funded by Angela. And given the nature of the problem that you’re solving, we also have some revenue. So couple of these help us but what we are doing right now is to raise about half a million, so that we are able to demonstrate the attainment of our revenues, and also build a product. I mean, right now we have a library of these modules, we want to put them in the cloud, they will always be on prem for now. But you want to make it on the cloud so that it’s more accessible, to get some training and support their product is a little more of our offering and accelerate this because I think the next few years is great, Greenfield opportunity in the space, especially this autonomous machine vision. We want to be the first in the industry to be delivering a truly autonomous machine vision platform for manufacturing so that anyone even with the basic knowledge should be able to utilize our local node platform and start With the visual inspection on the floor, that’s the vision every day deliver on that vision. And that’s what keeps us going.

Michael Waitze 15:05
And it was a really interesting animal. And I don’t mean animal meaning, meaning a living animal. I just mean, the way it runs its business. It’s kind of like a founder matching thing. Right? And I’m curious, did you go into antler with a co founder already? And with an idea, or did you just go in with the idea of building something and then came out with this? You know what I mean? I

Dushyant Verma 15:24
know a great question. Again, Rob. And I knew each other for the last six years, and we got this out of sales, we actually went to anther four and pitched the idea to them, we said, hey, why don’t you invest in us. So unlike the cohorts, that’s where the match with the founders, we came in with a product, we came in with a team, we came in with some revenues and some clients, they got us out, in fact, kudos to them, they bring discipline, they bring a lot more simplification of our business idea, especially when you talk the initial language, not many of them, understand into deep dive into technology, but how to make it palatable to them and how to make a vision much more streamlined. Easy. That’s what they’re integral. They’re an integral part of our team, we keep going to them for advisory. And the unique part about us is, I think we’ll be one of the few manufacturing and AI focused entities that we’ve invested in. Yeah, they feel that especially in Singapore, there was a couple of guys like svt, and entrepreneur first, they’re not doing a great job. But Angela, given it swing, both in Europe and across Asia, they will help us get us to investors, what is and become more than just like getting a VC fund. It was grow and accelerate, and deliver to our vision.

Michael Waitze 16:39
Yeah, I mean, I think what Apple has done, frankly, has just completely revolutionized the way investment gets done. And again, not just here, but all over the whole world. Rather, they’ve raised some more money for themselves, but they also have offices in like 30 countries now it’s kind of insane what they’ve done, you see and the team that are pretty amazing, I think, yeah,

Dushyant Verma 16:53
we interact with the team, and we find a certain quality of the founders, we tend to talk and live with a number of those guys working more around the data analytics side reporting side, because that’s like a huge deliverable that you provide to the manufacturers, something which is easily seen. So that’s really complimentary. So we find a lot of support from

Michael Waitze 17:12
it. That’s awesome. You mentioned earlier that you went in there with some revenue, which I think is great. I mean, if you building a company from scratch this idea, at least in my mind, right, sort of customer growth is the most important thing. But revenue growth doesn’t really matter to me is anathema. And I love the way you guys went in with real revenue and real customers. And can you talk to me about what you learned from the earliest implementations? Because when you build a product, right, no matter how much research you do, you think my customer is going to use my product in this way. And the result of their usage is going to be this output, right. But once you give it to them out in the wild, they’re probably going to do things with it that maybe you didn’t anticipate. So you learn some stuff, there were these learnings that then you got back and said, Oh, we can then build that into the product. And then our other customers can use it to makes the product more compelling was that like,

Dushyant Verma 17:54
my game? I call them things. Every day in our journey, like yesterday was new phase new. And tomorrow, we’ll also something new. So we always approached with this kind of learning for all our implementation. It’s a very simple thing, right? You take a manual process, understand, separate the quality inspectors classify those defects, put it in the system, and it gives it out for you. Right, there were a couple of great aha moments of epiphany for us. The first part was we realized that one of the customers that went up piano 1000 images, and did not know what to do with it. Now from our side, if you sat manually to analyze those images and annotate them, it would take us ages, we said that we need to add plus, but they wanted the other guys have failed. Can you deliver this to us said yes, we will. We will try. We went back, this was like a first base of a product where we were able to split the images into nano pixels and annotate each of them. We built this too. So in just about when either images, you got about a million data points, and that is what is required for the AI. Right, wow, we built this and we were done. And that is something that we still use today in terms of follow up. And that’s what makes our technology really. So we went out and we implements. The second part has always been more like learning where you’re trying to replace a manual or a human process, there is a certain quality like 95% 96%, you can always exceed that. That’s like okay, I’m able to do it much more faster. So process which was taking about four to 10 minutes, I’m able to do that under a minute, right. So it will just give me that result. And I’m here to deliver say about 99.9% efficiency because it’s a machine. But the real real part for us was when we when we were able to detect certain anomalies. A quality inspector would classify, say around five defects at max, we would constantly and we would classify any defect that’s coming and he wouldn’t know exactly this effect is happening right? But the moment of realization was when we found anomalies, right? So this was neither good nor bad. Somewhere in between when we asked this guy tell me what this is mostly The conditioning will say it’s good, or it’s bad. But there were certain instances that we would find it would not neither belong to the category. It was neither good nor interesting. And what we’ve done over them is to find a defect which was not here possible by human detection by because we have those five classification, which means there is a new kind of error, which is there in the system, which is completely identified. All these previous errors, would have some precedents, so known academics and scholars. But this new one requires a complete new investigation that really told us that there is further to then to our detection, detection is normally your operational data where you would take any of the assets, when the asset is getting manufactured, there is a lot of underlying influences, there is operational data, and the reliability data, also the machines etc. Can that be congregated together, find a much more meaningful instances of these defects. Those are the things that we take back and provide much more meaning we want to turn our quality inspectors into just people who provide numbers into more like data scientists so that they can provide meaningful inferences, actionable inferences down to the production line, this is one of

Michael Waitze 21:09
the things that technology at scale, but AI specifically is changing, right is that it also means you may want to you’re gonna impact the type of people that get hired to do those jobs, right. Because if you turn a quality engineer into a data scientist, that means they’re gonna play a much larger role in a whole bunch of other different processes, right? Like, they may change the entire manufacturing process, they may change the materials that get used, they may change the design, in a way, they may even change the sales, right? Because they’ll see all this data and think, How can I now take all this data that I’m getting, which I didn’t get before, and change the way the entire manufacturing process happens. But that also means you’re going to impact the type of person that needs to do that, right? Because if you can literally have a data scientist watching the production line, it’s gonna be a very different input than a mere quality engineer, not to minimize quality engineers. But you understand the point, it’s, it’s very different. No, absolutely,

Dushyant Verma 21:55
that’s like being vision, a convert these guys into much more meaningful, and that’s really coming up to Asia, I mean, now, to be more relevant in the next five to 10 years, you have to be a little more than just finding out those menial tasks, it’s going to be a very different world in next three or four years and preparing people for that, even if they don’t embrace it, they will have to embrace anyone to give them that input. However, it is more simplistic, more complex than what we say. Because a manufacturing process, especially for this midsize small manufacturers, it tends to be very siloed, their whole emphasis is on the batch and the production and the output, then actually, the underlying factor, the underlying factors play in when any sort of defect gets flagged by the parent, what we want to do is not just for this quality, but also for the management to have much more understanding about what their whole processes give them meaningful, insightful, actionable points than just a number, right? Otherwise, 99.9% is just a number and it can go up, it can go down, maybe it’ll change contextually. But what we want to give is a meaningful implication of any effect that’s going on that’s identified, so that they are much more adept at taking these actions. I’m sure there was some way before it was lying in a different silo. But our intention is always to congregate, bring them together in one platform. So that not just the quality inspector. For me, it’s down to the whole management, like bottom up so that there is more believability in the production process.

Michael Waitze 23:29
Yeah, look, it’s so exciting. The stuff that you’re building is just fascinating for me, right? The hardware, the software, the way all that stuff goes together, the way it impacts, industrial processes, manufacturing, and I think it’s going to feed into the rest of the business as well. You’re right. All these things are happening in separate silos, but the technology itself, yours is modular, but in a bigger modular way. You can connect all this stuff that’s happening there to other systems internally, and have way more impact. Look, I want to let you go. I want to thank you for doing this Dushyant Verma, the CEO of SmartViz. But I want you also to think about six months from now nine months from now, as you’re growing, to come back on the show and talk about what happened that you thought was gonna happen, the things that you didn’t anticipate how the business is growing and stuff like that. I really hope you come back. Thank you again for doing this today. Michael,

Dushyant Verma 24:10
I would love to do that. Because it’s an interesting to talk. It’s not just you, but talking. I realize oh, yeah, we’ve come this far. So I would love to share this. We would love to share with you again, a wonderful opportunity. Thank you so much for giving me

Michael Waitze 24:25
it’s my pleasure.

 

 

Latest Episodes:

EP 325 – Building a Marketplace and Democratizing M&A for SMEs – Marcus Yeung – Match.Asia and SEAbridge Partners

EP 325 – Building a Marketplace and Democratizing M&A for SMEs – Marcus Yeung – Match.Asia and SEAbridge Partners

So it’s actually very interesting how technology is taking a lot of the grunt work out of investment banking, which allows the real investment bankers to focus on the important stuff, which is, you know, the negotiations and the closing of the deals and more personal stuff. There is a good company in Japan, which has a good precedent called M&A Research Institute, which is a listed. And within it’s a listed company, it’s a similar business, and we’re following basically the same path as they have. And they closed 150 deals last year. And they made $60 million of revenue. And it’s a very profitable business model. – Marcus Yeung

read more
EP 324 – Creating a Canon for Progress: Empowering Global Entrepreneurs with Essential Knowledge – Tamara Winter – Commissioning Editor of Stripe Press

EP 324 – Creating a Canon for Progress: Empowering Global Entrepreneurs with Essential Knowledge – Tamara Winter – Commissioning Editor of Stripe Press

“Well, we think of this catalog as, ideally, a kind of cannon for progress, both in a very sort of explicit sense when it comes to, again, growing the GDP of the internet, making it much easier for people who have very ambitious ideas about a company they want to build to do that, and in an easier way than they might have to do if they were just trying to figure out everything themselves. You don’t need to reinvent the wheel.” – Tamara Winter

read more
EP 323 – Building Businesses and Breaking Barriers: A Global Entrepreneurial Odyssey – Meenah Tariq – CEO of Metric

EP 323 – Building Businesses and Breaking Barriers: A Global Entrepreneurial Odyssey – Meenah Tariq – CEO of Metric

In this episode of the Asia Tech Podcast, ⁠Meenah Tariq⁠, CEO of ⁠Metric⁠ and an experienced entrepreneur with deep roots in emerging markets, shared her journey and the lessons she’s learned along the way. Meenah discussed her early introduction to entrepreneurship in Pakistan, beginning with childhood ventures that sparked a lifelong passion for business.

read more