This episode examines artificial intelligence-driven optimization in manufacturing and CADDi's domain-specific intelligence. Yushiro Kato of CADDi, Inc. is co-founder and chief executive officer. Kato discusses CADDi-Drawer, a drawing-centric platform, vertical integration from marketplace to software, and how specialized intelligence and small language models accelerate product development in automotive, industrial and semiconductor supply chains, and they highlight practical examples of design reuse, automated risk flagging and supplier selection.
Key takeaways include CADDi's ability to reduce engineering lead times from weeks to hours by leveraging historical designs rather than generating designs from scratch, creating a downstream-to-upstream data flywheel, and addressing manufacturability risks early.
Kato describes encoding veteran domain expertise into agentic workflows that convert intelligence into measurable outcomes. John Furrier of theCUBE Research and analysts emphasize broader implications for physical AI and industry-scale acceleration.
Kato outlines CADDi's evolution from an Amazon-like marketplace with inspection centers to a software-focused company serving 22 countries and 900 employees, and shares a 10-year vision to accelerate physical innovation by 10x. The discussion highlights safety, access and cost as physical-layer constraints that stakeholders must solve to translate digital intelligence into real-world impact.
Forgot Password
Almost there!
We just sent you a verification email. Please verify your account to gain access to
theCUBE + NYSE Wired: Mixture of Experts Series. If you don’t think you received an email check your
spam folder.
Sign in to theCUBE + NYSE Wired: Mixture of Experts Series.
In order to sign in, enter the email address you used to registered for the event. Once completed, you will receive an email with a verification link. Open this link to automatically sign into the site.
Register For theCUBE + NYSE Wired: Mixture of Experts Series
Please fill out the information below. You will recieve an email with a verification link confirming your registration. Click the link to automatically sign into the site.
You’re almost there!
We just sent you a verification email. Please click the verification button in the email. Once your email address is verified, you will have full access to all event content for theCUBE + NYSE Wired: Mixture of Experts Series.
I want my badge and interests to be visible to all attendees.
Checking this box will display your presense on the attendees list, view your profile and allow other attendees to contact you via 1-1 chat. Read the Privacy Policy. At any time, you can choose to disable this preference.
Select your Interests!
add
Upload your photo
Uploading..
OR
Connect via Twitter
Connect via Linkedin
EDIT PASSWORD
Share
Forgot Password
Almost there!
We just sent you a verification email. Please verify your account to gain access to
theCUBE + NYSE Wired: Mixture of Experts Series. If you don’t think you received an email check your
spam folder.
Sign in to theCUBE + NYSE Wired: Mixture of Experts Series.
In order to sign in, enter the email address you used to registered for the event. Once completed, you will receive an email with a verification link. Open this link to automatically sign into the site.
Sign in to gain access to theCUBE + NYSE Wired: Mixture of Experts Series
Please sign in with LinkedIn to continue to theCUBE + NYSE Wired: Mixture of Experts Series. Signing in with LinkedIn ensures a professional environment.
Are you sure you want to remove access rights for this user?
Details
Manage Access
email address
Community Invitation
Yushiro Kato, CADDi, Inc
This episode examines artificial intelligence-driven optimization in manufacturing and CADDi's domain-specific intelligence. Yushiro Kato of CADDi, Inc. is co-founder and chief executive officer. Kato discusses CADDi-Drawer, a drawing-centric platform, vertical integration from marketplace to software, and how specialized intelligence and small language models accelerate product development in automotive, industrial and semiconductor supply chains, and they highlight practical examples of design reuse, automated risk flagging and supplier selection.
Key takeaways include CADDi's ability to reduce engineering lead times from weeks to hours by leveraging historical designs rather than generating designs from scratch, creating a downstream-to-upstream data flywheel, and addressing manufacturability risks early.
Kato describes encoding veteran domain expertise into agentic workflows that convert intelligence into measurable outcomes. John Furrier of theCUBE Research and analysts emphasize broader implications for physical AI and industry-scale acceleration.
Kato outlines CADDi's evolution from an Amazon-like marketplace with inspection centers to a software-focused company serving 22 countries and 900 employees, and shares a 10-year vision to accelerate physical innovation by 10x. The discussion highlights safety, access and cost as physical-layer constraints that stakeholders must solve to translate digital intelligence into real-world impact.
play_circle_outlineAI Manufacturing Platform with Small Specialized Language Models Reusing Designs, Flagging Risks, Cutting Engineering Lead Time to Hours
replyShare Clip
play_circle_outlineReuse-first design, automated CAD/drawing checks, and procurement automation for supplier selection and cost benchmarking
replyShare Clip
play_circle_outlineCADDi Pivot to Software: Launching CADDi Drawer, 900 Employees Serving Discrete Manufacturing in 22 Countries
replyShare Clip
play_circle_outlineData flywheel connecting downstream manufacturing feedback upstream to designers
>> (INTRO)Hello, I'm John Furrier, host of theCUBE. We are here in theCUBE's NYSE Wired studio. Of course, we are Palo Alto studio connecting Silicon Valley to Wall Street. This is our Mixture of Experts series by the NYSE Wired program and community. We got Yushiro Kato, the co-founder and CEO of CADDi Inc., doing really cool stuff with AI in a very specific domain. Again, highlighting how the domains are lighting up with intelligence. They're doing it kind of different, but the same. Thanks for coming on theCUBE. Appreciate
Yushiro Kato
>> it.Thank you for having
John Furrier
>> me.So first of all, explain what you guys do, because I think this is a great use case that validates what we've been saying for 2 years, small language models first, specialized intelligence is coming on the scene. You guys are harnessing that intelligence. Explain what you guys do.
Yushiro Kato
>> Yeah, absolutely. So CADDi is a vertically integrated platform for manufacturing, and our AI automatically ingests the data, like past data of manufacturing companies, which is scattered across systems and structures, and analyze the data with our proprietary AI. As you said, small language model or small, a specific AI and then puts AI to work on it. So basically, let's say a hardware engineer, if they receive a requirement from the customer, our AI starts to work on it and then reads it and understands it and then finds the closest designs that they've ever made and which can be reused. And then it automatically finds the changes required and flags risks associated with the changes. So basically it automatically dramatically shortens the lead time of product development. From 2 weeks, several weeks to several hours.
John Furrier
>> You know, one of the hot things right now is Chief of Staff products. I play with them all the time. I don't let it get access to all my stuff, but I'll give it access to my email. Okay. My email gets out there. Okay, whatever. But that talks to that domain assistant. In your example you just gave, that's essentially a coworker. That's like having a staff. This is where you start to see action being taken, not just results, search results.
Yushiro Kato
>> Exactly.
John Furrier
>> This is where we start to see the work. Take us through that value proposition and compare it to what would have been needed to harmonize data across disparate systems. Supply chain manufacturing, they're dealing with stuff all the time.
Yushiro Kato
>> Yeah,
John Furrier
>> in general. But now you have an engineer, a requirement that's specific. What would it have been like before and after?
Yushiro Kato
>> Yeah. So basically engineers or procurement, we work with a lot of people across the value chain from engineers, procurement, production engineering, quality assurance, etc. They of course are an expert using Excel files and emails and phone calls. Right. And what we do is to, as you said, agents are now like members, right? Like members. So they complete tasks and then report it to the real users and then they make decisions, right? So it's not a nice-to-have tool. It's literally changing the way people work and complete tasks.
John Furrier
>> I mean, no one gives it up once they touch it. Explain that psychology of the outcome, if you could. Take us, walk us through an example.
Yushiro Kato
>> Okay. So I talked about an engineering example, right? So in the conceptual design phase, again, AI reads the requirement, understands what they need to build and then finds the closest drawings. By the way, this is very important, like manufacturing. In manufacturing, a risk is a change. Change. So change is a risk. So there are a lot of generative AIs, right, that generate something from scratch, which is great. But actually, you know, generating something has a huge risk. Associated with it. So basically our idea is why don't you just leverage what you have built in the last 100 years, right? And then, leverage that and then identify what needs to be improved from there, not from scratch. And anyways, so engineers start from the kind of starting design. And then once they finished the detailed design, like on drawings and 3D CAD models, our AI automatically checks it and flags risks, like potential risks in manufacturing and in quality. And then so that engineers can realize it before it goes through the production, like real production. And in terms of procurement, we help procurement people to reduce the cost and reduce the time to select suppliers. So our product automatically selects suppliers for our procurement offices and then it gives cost benchmark based on the past parts.
John Furrier
>> I love this example because engineers are skeptical. There's a company out there called Harvey that does AI for lawyers, and they were a huge success. Lawyers are skeptical, but they know their work. They know what good looks like so they can see value super fast. Even CFOs on the finance side, we're seeing that vertical light up with AI because they know what good is. Explain the phenomenon in your vertical because it's the same. What would be the work required pre-AI? What would they have to do and what's the time difference and what risks come out of that? Take us through that before and after in detail. What would it be like?
Yushiro Kato
>> Yeah, so basically all the knowledge is in human heads now, right? Veterans know what happened to similar drawings like 5 years ago and they can immediately identify, oh, this tolerance is too tight. So you just like, we should loosen it, whatever that is, right? And then, so it was all manual and veterans knowledge oriented. Now younger, even younger engineers without any experience at the company, without understanding the company's history, they can start working on the product literally from day one. So we are shortening a lead time again from several weeks to hours for engineering. They basically don't need to ask what happened in the past to senior engineers and they can just search it within their asset. And then so like
John Furrier
>> they have history, they know when they see a recommendation come in or an error, they know that if I had missed that, they would have known the consequence. Yeah, they can quantify that.
Yushiro Kato
>> Yeah, exactly. And before AI, there were a lot of reworks. Like rework is one of the biggest bottleneck to shorten lead time. So let's take an example of automotive. Automotive companies, one car consists of 30,000 parts and it takes 4 years to build a car, from planning to really build a car. And then for every part, it generally there are 10 to 20 reworks. So sometimes design is fixed and then creates some prototype and then finds errors on drawings and get back to the drawing and
John Furrier
>> design.Yeah, redo
Yushiro Kato
>> it.And then redo it over and over again, right? So what if you know potential risks at the early stage and then mitigate the risk? You can
>> cost.Yeah, it's a cost and it's time, right? It's really a lot. It takes a lot of time and cost.
John Furrier
>> All right, talk about the impact to the industry because manufacturing, we're seeing people, I was even talking to some OpenAI people about they're using their intelligence to build their own chip, Jalapeño. There's other companies out there that have encapsulated accumulated domain expertise in certain areas and they have their own language models or foundation models, not just language, but physics, semiconductors. We're seeing that similar things are happening in manufacturing. So I see the value. What's it like with the customer? What's the motion? You just walk in and say, hey, I got a tool, it's going to be great. Take us through that. is there a blocker? Is there skepticism? What's the motion? Do they do a trial? Because I can imagine they're busy and they— yeah, they're not just big R&D departments. They want immediacy.
Yushiro Kato
>> Yeah, I think hardware engineers and people in manufacturing by nature are skeptical about software, honestly speaking. And honestly, until several years ago, software and SaaS were kind of scratching the surface. Yeah, it's good, but to improve 2% of the HR efficiency or 2% of accounting efficiency. It doesn't transform the company, right? And we actually started our company as a platform provider. So we provided CADDi Manufacturing, which is an Amazon-like marketplace for custom-made parts. And we were liable for quality, cost, and delivery ourselves. So we had like more than 10 inspection centers. So we were manufacturers. We were skeptical.
John Furrier
>> You lived it.
Yushiro Kato
>> Yeah, exactly. About the off-the-shelf softwares because there are no system integrators or no system providers who really have an experience of doing manufacturing, right? So we needed to develop more than 20 applications internally to streamline our own business first. And then customers started to ask us to productize the technologies. So we know the pain points, we know how manufacturing works.
John Furrier
>> Yeah, you've lived it, you built it. Yeah, exactly. For engineers, by engineers. And by the way, that's credibility. Talk about the company. Give us a scope of the size. What you guys are working on, things you're pioneering, areas you're improving. Give an overview.
Yushiro Kato
>> Yeah. Okay. So we, CADDi, I started the company with a co-founder, Aki, who used to be an Apple engineer at Apple in 2017. And again, the first business was CADDi Manufacturing, which is a marketplace. And then we built 26 applications internally. And then again, customers started to ask us to productize one of the core technologies, which is design analysis, like drawing analysis technology. So we launched CADDi Drawer 4 years ago, which is a drawing-centric data platform. And then 2 years ago, we pivoted from manufacturing business to software. We had 2 businesses in parallel, but I kind of was spread too thin. So, we started to focus on the software. So now we have 900 people across the globe. We have customers in 22 countries, manufacturers, and
John Furrier
>> they do supply chain procurement. They're optimizing for that.
Yushiro Kato
>> Yes.
John Furrier
>> Okay.
Yushiro Kato
>> And so we're primarily focusing on discrete manufacturing. So automotive, industrial tools, semiconductor manufacturing machinery like packaging machine. Very Apple-like. Yeah.
John Furrier
>> I mean, there's a lot of engineering there. This is the beautiful thing about this is that the domain expertise out there— talk to people that are watching and share your opinion on how important it's taking that domain expertise either on the head, in the heads of the engineers or founders or leaders and encoding that into the intelligence. Talk about how to think about it, what's the right approach because you're starting to see the agentic side of the business where intelligence is action-oriented, doing work and there's an outcome that can be quantified where that's where the value is. Industrial capital.
Yushiro Kato
>> Yeah. So domain expertise is, I think, really important. So for example, we had, as I said, more than 10 inspection centers, physically, we were physically inspecting parts, right? I myself personally deburred a lot of stainless parts and burrs. And so, I know the kind of firsthand— I have firsthand experiences, right? And then, for example, startups are here to change the industry, right? And you need to basically change the conventional wisdom. So for example, quality, of course, quality is an issue, everybody knows that. But we had 10 inspection centers and captured all the data of quality. And actually more than 50% of all the quality defects happened because of the quality of drawings and 3D CAD models, not because of the poor manufacturing or because of suppliers. It's actually, it was like lying in
John Furrier
>> theroot cause
Yushiro Kato
>> withthe drawings and like designs. So, we were a procurement parts provider at the time, but we couldn't change that design, right? Like we— Yeah, you were downstream. Yeah, exactly. We gave feedback to engineers, but engineers, for engineers, there's no incentive.
John Furrier
>> Yeah.Thanks for the advice.
Yushiro Kato
>> I'll work on that next time. Exactly. If that is fixed. But after launching CADDi Drawer, engineers naturally started to fix the changes when they were working on the new parts because they can understand, oh, these are the similar parts they have manufactured and these are the quality defects happened to those similar parts. So, you know, why—
John Furrier
>> you essentially created the workflow that was disparate.
Yushiro Kato
>> Yeah.
John Furrier
>> And you unified it. Yeah. With AI. Yeah. And you went right to the point where the problems were.
Yushiro Kato
>> Yes.
John Furrier
>> Versus just writing an email or giving feedback.
Yushiro Kato
>> Yes, absolutely. So now value chain is disconnected. So engineers, of course, really do not care about manufacturability, right? But manufacturability is really important. But, manufacturers know, if they can manufacture it, how much the cost is, but it's too late, right? So what we are doing is to create what we call data flywheel from downstream to upstream so that people at the early stage can really mitigate the risk, again, while they are working.
John Furrier
>> All right. Well, what's next for you guys? Obviously, the market's booming. You're starting to see manufacturing come back. There's a huge demand in all verticals, whether it's healthcare, manufacturing, semiconductors. A lot more conventional wisdom being broken down and rebuilt. What's next for you guys?
Yushiro Kato
>> Yeah, so we have a vision, like 10-year vision, which is accelerate physical innovation by 10x. What I mean by that is that AI, everyone is talking about AI. AI is exciting. Right. But when AI is automatically solving mathematical theorems, for example, and the number of theories and ideas coming out of it is kind of exponentially growing. But when we look at the physical world we live in, how much has it changed? Like homes and commutes and hospitals and cars? Not much.
John Furrier
>> Yeah. we need help.
Yushiro Kato
>> Yeah, exactly. And why is that? Because of course, in order to change the physical world, someone needs to build something for the physical world. That's the physical world, right? And but material, like what we call materialization, and it takes time. Again, now it takes 4 years to build a car, but even 30 years ago in 1995, it took 4 years to build a car. It didn't speed up that much, right? So unlike software. So what we are doing is to again, like a lot of one critical issue here is that In software, software developers use, actually write less than 10% of the codes. They actually fetch the data from libraries, SDKs, and frameworks that someone else has created. In manufacturing, more than 80% of the designs are created from scratch. So what we are doing is to create a library, it's industry-wide library and company-wide library. And help them to leverage what's very
John Furrier
>> NVIDIA-liketoo. They're really donating a lot of that IP as open code and digital twins to assist people as a service to move the industry forward. Okay, final question. So I got you here because of Mixture of Experts, we could probably get you in the Physical AI series because one of the things that we're watching now, it's our second year into, is one of our popular series called Physical AI and Robotics is this notion of intelligence in the real world. Just say real world. That's not just robotics and cars. We know that we love those. Who doesn't love autonomous cars? We're talking about buildings, cities, and towns. We operate as humans in a physical world. The digital is coming together. We have devices. This total blurring of the lines. This is the convergence. You're smiling. Share your vision because again, this span. This is happening. Timeframe you just laid out a little bit longer, but it's happening. Take us through that vision because this really changes everything. Clothes we wear, potentially the homes we live in, cars we drive. First one to check the box is automobiles. Manufacturing, easy use case, great way to enter the market. What's beyond manufacturing?
Yushiro Kato
>> Yeah, so the bottleneck I think is shifting from intelligence layer. Again, AI has kind of surpassed the human intelligence already, but Bottleneck is shifting from intelligence layer to physical layer. So we need to build something, right? And why is that important? Because we categorized what we call physical limitations to 14 categories, one of which is safety. For example, even if software is super smart, if an aircraft goes down, people die. It's a real physical issue, right? I lost one, a friend, best friend of mine because of an accident in Miami. It's a physical limit. And that's one. And also, for example, an access to physical things. So I love traveling. I traveled to more than 30 countries and I saw a lot of poverty, right? And more than 4 billion people on the earth do not have any access to a car. And of course it's because of poverty, but it's also because of the cost of cars, right? If that is less than $1,000, actually more than 80% of the 4 billion people can have access to a car. And that's what SpaceX changed, right? Like SpaceX, like space shuttle, the cost was $100 like 25 years ago and now it's $3 for them. They can reuse it. Yeah. And the cost of they had
John Furrier
>> leverage.
Yushiro Kato
>> Yeah. So they are 14. Air pollution is one thing and, you know, travel
John Furrier
>> and it's an acceleration you're getting at. And that's like Moore's Law applied to the real world. Yeah. You're starting to see economies of scale, lower costs, better quality.
Yushiro Kato
>> But someone needs to do that, right? And people are spending more than 99% of time, in manufacturing, on repetitive tasks, somewhat recurring, somewhat similar tasks. So why don't you just leverage historical knowledge and shorten that lead time and shorten that time and then let them focus on something that is more important?
John Furrier
>> Yushiro, thank you for coming on. I really appreciate it. And I'm sorry to hear about your friend. The physical world is dangerous, as is AI if you don't rein it in. But I really appreciate your vision and the physical world is coming fast. Yeah, thanks for coming on our Mixture of Experts and potentially our physical AI. We'll get you back in. Great vision. I'm John Furrier sharing the data here as intelligence injects into the real world. You start seeing changes we've never seen before, not just cars, how we operate as humans in society, lifting people out of poverty potentially. The world is about to change significantly. The AI wave is going to do that. Thanks for watching.
>> (INTRO)Hello, I'm John Furrier, host of theCUBE. We are here in theCUBE's NYSE Wired studio. Of course, we are Palo Alto studio connecting Silicon Valley to Wall Street. This is our Mixture of Experts series by the NYSE Wired program and community. We got Yushiro Kato, the co-founder and CEO of CADDi Inc., doing really cool stuff with AI in a very specific domain. Again, highlighting how the domains are lighting up with intelligence. They're doing it kind of different, but the same. Thanks for coming on theCUBE. Appreciate
Yushiro Kato
>> it.Thank you for having
John Furrier
>> me.So first of all, explain what you guys do, because I think this is a great use case that validates what we've been saying for 2 years, small language models first, specialized intelligence is coming on the scene. You guys are harnessing that intelligence. Explain what you guys do.
Yushiro Kato
>> Yeah, absolutely. So CADDi is a vertically integrated platform for manufacturing, and our AI automatically ingests the data, like past data of manufacturing companies, which is scattered across systems and structures, and analyze the data with our proprietary AI. As you said, small language model or small, a specific AI and then puts AI to work on it. So basically, let's say a hardware engineer, if they receive a requirement from the customer, our AI starts to work on it and then reads it and understands it and then finds the closest designs that they've ever made and which can be reused. And then it automatically finds the changes required and flags risks associated with the changes. So basically it automatically dramatically shortens the lead time of product development. From 2 weeks, several weeks to several hours.
John Furrier
>> You know, one of the hot things right now is Chief of Staff products. I play with them all the time. I don't let it get access to all my stuff, but I'll give it access to my email. Okay. My email gets out there. Okay, whatever. But that talks to that domain assistant. In your example you just gave, that's essentially a coworker. That's like having a staff. This is where you start to see action being taken, not just results, search results.
Yushiro Kato
>> Exactly.
John Furrier
>> This is where we start to see the work. Take us through that value proposition and compare it to what would have been needed to harmonize data across disparate systems. Supply chain manufacturing, they're dealing with stuff all the time.
Yushiro Kato
>> Yeah,
John Furrier
>> in general. But now you have an engineer, a requirement that's specific. What would it have been like before and after?
Yushiro Kato
>> Yeah. So basically engineers or procurement, we work with a lot of people across the value chain from engineers, procurement, production engineering, quality assurance, etc. They of course are an expert using Excel files and emails and phone calls. Right. And what we do is to, as you said, agents are now like members, right? Like members. So they complete tasks and then report it to the real users and then they make decisions, right? So it's not a nice-to-have tool. It's literally changing the way people work and complete tasks.
John Furrier
>> I mean, no one gives it up once they touch it. Explain that psychology of the outcome, if you could. Take us, walk us through an example.
Yushiro Kato
>> Okay. So I talked about an engineering example, right? So in the conceptual design phase, again, AI reads the requirement, understands what they need to build and then finds the closest drawings. By the way, this is very important, like manufacturing. In manufacturing, a risk is a change. Change. So change is a risk. So there are a lot of generative AIs, right, that generate something from scratch, which is great. But actually, you know, generating something has a huge risk. Associated with it. So basically our idea is why don't you just leverage what you have built in the last 100 years, right? And then, leverage that and then identify what needs to be improved from there, not from scratch. And anyways, so engineers start from the kind of starting design. And then once they finished the detailed design, like on drawings and 3D CAD models, our AI automatically checks it and flags risks, like potential risks in manufacturing and in quality. And then so that engineers can realize it before it goes through the production, like real production. And in terms of procurement, we help procurement people to reduce the cost and reduce the time to select suppliers. So our product automatically selects suppliers for our procurement offices and then it gives cost benchmark based on the past parts.
John Furrier
>> I love this example because engineers are skeptical. There's a company out there called Harvey that does AI for lawyers, and they were a huge success. Lawyers are skeptical, but they know their work. They know what good looks like so they can see value super fast. Even CFOs on the finance side, we're seeing that vertical light up with AI because they know what good is. Explain the phenomenon in your vertical because it's the same. What would be the work required pre-AI? What would they have to do and what's the time difference and what risks come out of that? Take us through that before and after in detail. What would it be like?
Yushiro Kato
>> Yeah, so basically all the knowledge is in human heads now, right? Veterans know what happened to similar drawings like 5 years ago and they can immediately identify, oh, this tolerance is too tight. So you just like, we should loosen it, whatever that is, right? And then, so it was all manual and veterans knowledge oriented. Now younger, even younger engineers without any experience at the company, without understanding the company's history, they can start working on the product literally from day one. So we are shortening a lead time again from several weeks to hours for engineering. They basically don't need to ask what happened in the past to senior engineers and they can just search it within their asset. And then so like
John Furrier
>> they have history, they know when they see a recommendation come in or an error, they know that if I had missed that, they would have known the consequence. Yeah, they can quantify that.
Yushiro Kato
>> Yeah, exactly. And before AI, there were a lot of reworks. Like rework is one of the biggest bottleneck to shorten lead time. So let's take an example of automotive. Automotive companies, one car consists of 30,000 parts and it takes 4 years to build a car, from planning to really build a car. And then for every part, it generally there are 10 to 20 reworks. So sometimes design is fixed and then creates some prototype and then finds errors on drawings and get back to the drawing and
John Furrier
>> design.Yeah, redo
Yushiro Kato
>> it.And then redo it over and over again, right? So what if you know potential risks at the early stage and then mitigate the risk? You can
>> cost.Yeah, it's a cost and it's time, right? It's really a lot. It takes a lot of time and cost.
John Furrier
>> All right, talk about the impact to the industry because manufacturing, we're seeing people, I was even talking to some OpenAI people about they're using their intelligence to build their own chip, Jalapeño. There's other companies out there that have encapsulated accumulated domain expertise in certain areas and they have their own language models or foundation models, not just language, but physics, semiconductors. We're seeing that similar things are happening in manufacturing. So I see the value. What's it like with the customer? What's the motion? You just walk in and say, hey, I got a tool, it's going to be great. Take us through that. is there a blocker? Is there skepticism? What's the motion? Do they do a trial? Because I can imagine they're busy and they— yeah, they're not just big R&D departments. They want immediacy.
Yushiro Kato
>> Yeah, I think hardware engineers and people in manufacturing by nature are skeptical about software, honestly speaking. And honestly, until several years ago, software and SaaS were kind of scratching the surface. Yeah, it's good, but to improve 2% of the HR efficiency or 2% of accounting efficiency. It doesn't transform the company, right? And we actually started our company as a platform provider. So we provided CADDi Manufacturing, which is an Amazon-like marketplace for custom-made parts. And we were liable for quality, cost, and delivery ourselves. So we had like more than 10 inspection centers. So we were manufacturers. We were skeptical.
John Furrier
>> You lived it.
Yushiro Kato
>> Yeah, exactly. About the off-the-shelf softwares because there are no system integrators or no system providers who really have an experience of doing manufacturing, right? So we needed to develop more than 20 applications internally to streamline our own business first. And then customers started to ask us to productize the technologies. So we know the pain points, we know how manufacturing works.
John Furrier
>> Yeah, you've lived it, you built it. Yeah, exactly. For engineers, by engineers. And by the way, that's credibility. Talk about the company. Give us a scope of the size. What you guys are working on, things you're pioneering, areas you're improving. Give an overview.
Yushiro Kato
>> Yeah. Okay. So we, CADDi, I started the company with a co-founder, Aki, who used to be an Apple engineer at Apple in 2017. And again, the first business was CADDi Manufacturing, which is a marketplace. And then we built 26 applications internally. And then again, customers started to ask us to productize one of the core technologies, which is design analysis, like drawing analysis technology. So we launched CADDi Drawer 4 years ago, which is a drawing-centric data platform. And then 2 years ago, we pivoted from manufacturing business to software. We had 2 businesses in parallel, but I kind of was spread too thin. So, we started to focus on the software. So now we have 900 people across the globe. We have customers in 22 countries, manufacturers, and
John Furrier
>> they do supply chain procurement. They're optimizing for that.
Yushiro Kato
>> Yes.
John Furrier
>> Okay.
Yushiro Kato
>> And so we're primarily focusing on discrete manufacturing. So automotive, industrial tools, semiconductor manufacturing machinery like packaging machine. Very Apple-like. Yeah.
John Furrier
>> I mean, there's a lot of engineering there. This is the beautiful thing about this is that the domain expertise out there— talk to people that are watching and share your opinion on how important it's taking that domain expertise either on the head, in the heads of the engineers or founders or leaders and encoding that into the intelligence. Talk about how to think about it, what's the right approach because you're starting to see the agentic side of the business where intelligence is action-oriented, doing work and there's an outcome that can be quantified where that's where the value is. Industrial capital.
Yushiro Kato
>> Yeah. So domain expertise is, I think, really important. So for example, we had, as I said, more than 10 inspection centers, physically, we were physically inspecting parts, right? I myself personally deburred a lot of stainless parts and burrs. And so, I know the kind of firsthand— I have firsthand experiences, right? And then, for example, startups are here to change the industry, right? And you need to basically change the conventional wisdom. So for example, quality, of course, quality is an issue, everybody knows that. But we had 10 inspection centers and captured all the data of quality. And actually more than 50% of all the quality defects happened because of the quality of drawings and 3D CAD models, not because of the poor manufacturing or because of suppliers. It's actually, it was like lying in
John Furrier
>> theroot cause
Yushiro Kato
>> withthe drawings and like designs. So, we were a procurement parts provider at the time, but we couldn't change that design, right? Like we— Yeah, you were downstream. Yeah, exactly. We gave feedback to engineers, but engineers, for engineers, there's no incentive.
John Furrier
>> Yeah.Thanks for the advice.
Yushiro Kato
>> I'll work on that next time. Exactly. If that is fixed. But after launching CADDi Drawer, engineers naturally started to fix the changes when they were working on the new parts because they can understand, oh, these are the similar parts they have manufactured and these are the quality defects happened to those similar parts. So, you know, why—
John Furrier
>> you essentially created the workflow that was disparate.
Yushiro Kato
>> Yeah.
John Furrier
>> And you unified it. Yeah. With AI. Yeah. And you went right to the point where the problems were.
Yushiro Kato
>> Yes.
John Furrier
>> Versus just writing an email or giving feedback.
Yushiro Kato
>> Yes, absolutely. So now value chain is disconnected. So engineers, of course, really do not care about manufacturability, right? But manufacturability is really important. But, manufacturers know, if they can manufacture it, how much the cost is, but it's too late, right? So what we are doing is to create what we call data flywheel from downstream to upstream so that people at the early stage can really mitigate the risk, again, while they are working.
John Furrier
>> All right. Well, what's next for you guys? Obviously, the market's booming. You're starting to see manufacturing come back. There's a huge demand in all verticals, whether it's healthcare, manufacturing, semiconductors. A lot more conventional wisdom being broken down and rebuilt. What's next for you guys?
Yushiro Kato
>> Yeah, so we have a vision, like 10-year vision, which is accelerate physical innovation by 10x. What I mean by that is that AI, everyone is talking about AI. AI is exciting. Right. But when AI is automatically solving mathematical theorems, for example, and the number of theories and ideas coming out of it is kind of exponentially growing. But when we look at the physical world we live in, how much has it changed? Like homes and commutes and hospitals and cars? Not much.
John Furrier
>> Yeah. we need help.
Yushiro Kato
>> Yeah, exactly. And why is that? Because of course, in order to change the physical world, someone needs to build something for the physical world. That's the physical world, right? And but material, like what we call materialization, and it takes time. Again, now it takes 4 years to build a car, but even 30 years ago in 1995, it took 4 years to build a car. It didn't speed up that much, right? So unlike software. So what we are doing is to again, like a lot of one critical issue here is that In software, software developers use, actually write less than 10% of the codes. They actually fetch the data from libraries, SDKs, and frameworks that someone else has created. In manufacturing, more than 80% of the designs are created from scratch. So what we are doing is to create a library, it's industry-wide library and company-wide library. And help them to leverage what's very
John Furrier
>> NVIDIA-liketoo. They're really donating a lot of that IP as open code and digital twins to assist people as a service to move the industry forward. Okay, final question. So I got you here because of Mixture of Experts, we could probably get you in the Physical AI series because one of the things that we're watching now, it's our second year into, is one of our popular series called Physical AI and Robotics is this notion of intelligence in the real world. Just say real world. That's not just robotics and cars. We know that we love those. Who doesn't love autonomous cars? We're talking about buildings, cities, and towns. We operate as humans in a physical world. The digital is coming together. We have devices. This total blurring of the lines. This is the convergence. You're smiling. Share your vision because again, this span. This is happening. Timeframe you just laid out a little bit longer, but it's happening. Take us through that vision because this really changes everything. Clothes we wear, potentially the homes we live in, cars we drive. First one to check the box is automobiles. Manufacturing, easy use case, great way to enter the market. What's beyond manufacturing?
Yushiro Kato
>> Yeah, so the bottleneck I think is shifting from intelligence layer. Again, AI has kind of surpassed the human intelligence already, but Bottleneck is shifting from intelligence layer to physical layer. So we need to build something, right? And why is that important? Because we categorized what we call physical limitations to 14 categories, one of which is safety. For example, even if software is super smart, if an aircraft goes down, people die. It's a real physical issue, right? I lost one, a friend, best friend of mine because of an accident in Miami. It's a physical limit. And that's one. And also, for example, an access to physical things. So I love traveling. I traveled to more than 30 countries and I saw a lot of poverty, right? And more than 4 billion people on the earth do not have any access to a car. And of course it's because of poverty, but it's also because of the cost of cars, right? If that is less than $1,000, actually more than 80% of the 4 billion people can have access to a car. And that's what SpaceX changed, right? Like SpaceX, like space shuttle, the cost was $100 like 25 years ago and now it's $3 for them. They can reuse it. Yeah. And the cost of they had
John Furrier
>> leverage.
Yushiro Kato
>> Yeah. So they are 14. Air pollution is one thing and, you know, travel
John Furrier
>> and it's an acceleration you're getting at. And that's like Moore's Law applied to the real world. Yeah. You're starting to see economies of scale, lower costs, better quality.
Yushiro Kato
>> But someone needs to do that, right? And people are spending more than 99% of time, in manufacturing, on repetitive tasks, somewhat recurring, somewhat similar tasks. So why don't you just leverage historical knowledge and shorten that lead time and shorten that time and then let them focus on something that is more important?
John Furrier
>> Yushiro, thank you for coming on. I really appreciate it. And I'm sorry to hear about your friend. The physical world is dangerous, as is AI if you don't rein it in. But I really appreciate your vision and the physical world is coming fast. Yeah, thanks for coming on our Mixture of Experts and potentially our physical AI. We'll get you back in. Great vision. I'm John Furrier sharing the data here as intelligence injects into the real world. You start seeing changes we've never seen before, not just cars, how we operate as humans in society, lifting people out of poverty potentially. The world is about to change significantly. The AI wave is going to do that. Thanks for watching.
John Furrier