Nineteen Year Old Parth Mehta Reinvents CAD with AI

We talk with Makistry founder Parth Meta about turning plain English prompts into parametric CAD and why a structured AI “brainstorm” can speed design without losing engineering control. We dig into standards-aware reasoning, exports, limitations, and the roadmap for assemblies and 2D-to-3D.

• Why CAD still slows real projects
• Text to parametric models through a guided brainstorm
• Using RAG to ground standards like M4 and hole specs
• Open Cascade kernel, STEP exports, in-browser visualization
• Measuring and parameter checks for trust and repeatability
• Limits on complexity and plans for sketches to 3D
• Assemblies, mates, and constraints on the roadmap
• Moving beyond copilots to AI-native design and CAM
• Balancing startup work with school and sport
• Making geometry organic yet manufacturable

Transcript

Meet Parth And Makeistry

Roopinder 0:00

My guest today is Parth Meta. Parth is a founder of Makistry. You say Make Istree or Mechistree? Make history. Makistry is a AI-powered CAD program. And uh we're very interested in hearing about it because I've long said CAD is very difficult to learn for most people. And uh it took me 30 years to realize that. But here's someone who's learned it right away at a very early age. Barth, welcome. Welcome to the uh show. Thank you. Thank you. Yeah. Yeah. Would you tell us a little bit about yourself and what led you to make Makistry?

CAD Pain Points And The Spark

From Class Project To Prototype

Parth 0:48

Yeah, so yeah, I'm part. I'm a student at UT Austin, um, studying mechanical engineering and economics. Yeah. And the story behind Makistry was so I kind of had the idea for this close to six or seven months ago, um, where I was in a design class as a mechanical engineer, you have to do like introduced design classes. And what ended up so we had to make a battle robot from scratch. And what ended up happening was our team and most of the other teams in our class kind of spent eight to ten weeks out of the 12 weeks that we had that semester to build the project, just working on CAD models, creating CAD models, iterating on them, kind of running simulations, figuring out what's going on, and then kind of iterating on them again. So that's when it kind of hit me like CAD is just taking too long. And there's got to be a better way to do this because I don't want, I don't want to be spending my time kind of working on CAD, creating CAD models. I want to be thinking about new ways that I could design or innovative ways that I could approach and solve a problem. So that's kind of where the Spark for Makistry came up. I was basically looking for a way to make CAD easier and make CAD faster. And yeah, I kind of dove in, did some research. And once the summer came, I kind of dove into it. I was doing an internship, but it was a little boring. So I was like, I don't want to do this. I'll put a minimal effort on that and focus on something that I really want to build. And I kind of spent that summer doing a lot of research, doing a lot of learning, and built Makistry to a point where it was a working prototype off a text-to-CAD kind of tool. Um, and before that, like during the development process, I also ran a survey. Um we got responses from over 200 makers just to see if other makers also had the same uh feelings about CAD. And what we found was they did. They had similar complaints that I did. It was just took took too much time, the learning curves were really steep. Um, the CAD tools were built decades ago and hadn't changed in years. So they were looking for better ways to make CAD models. And most AI tools at that point were creating meshes. So it was just a text-to-mesh kind of tool, which wasn't usable in engineering workflows. So that's kind of where I found that little gap where AI for parametric CAD wasn't really a big thing back then. So that's kind of like the niche I targeted and built a working prototype over the summer where you could convert text prompts into parametric CAD models. And yeah, and then once the semester started again, um I figured that I wouldn't be able to continue building with just these two hands. So I there's an organization here at UT called Texas Momentum. So I applied there, they accepted me. They basically helped student startups get students to help um with their projects and help build stuff up. So we got a team of engineers from them, and we basically converted the prototype into a full-fledged kind of web app, and we launched that as our beta in October. And since then, the traction's been really good. Um, things are coming along really well. So yeah, that's kind of the story. And the name Makistry is actually make with artistry. So we're helping makers make with artistry.

Roopinder 4:14

Very good, very good. Very good intro. I want to go into each of those things that you mentioned. It's very interesting to me how, like I said, how you how you realize CAD was hard because you know that's been my that's been my rant for uh several years now. And I have um I've mastered several CAD programs and and hacked my way through several others, and uh maybe you know, like all uh all of them, all the major CAD players I've had some association with. And what I find is most users will not confess to that, will not confess to it being hard. They will uh they will say, Oh, it's easy to use. And I my my uh retort then is usually well it's easy to use if you know how to use it. And that's you know, that that to me is like, oh, for anybody else, it takes a long time. And even now, yeah, even now, even after being a veteran of CAD systems, you probably use SOLIDWORKS if I'm correct, right?

Parth 5:14

Yeah, yeah, SOLIDWORKS on shape feed and six feet. I tried everything to see how it works, yeah.

Surveys Validate The Need

Roopinder 5:19

Right. And uh so still to this day, when I want to make something, I'm a mechanical engineer, uh mechanical engineers, right? To this day, I I want to, if I want to make something, I'll make stuff in my shop. I'm not a real manufacturer of any kind, but you know, I like to work in my shop, I do woodworking. But before I make something, I don't use CAD. I don't use CAD. I should, I feel guilty. Hey, I don't use CAD. And I for after a while I had to had to do some soul searching. Yeah, when I using CAD when all this time I'm promoting CAD, but I don't use it because and I realize it's too damn difficult even for me to use. It takes me too long. I can go into the shop with a hand sketch, which by the way, I'm using a freeform in my Apple tablet because I it's a very nice interface. You know, I have a pen and a tablet, I can sketch on it. Uh, you know, it's not perfect, I can, and it bothers me a little bit, but I could go in the shop with that design and I'm off to the races. I'm starting to make things, right? Yeah, so that's the world we're in, right? So credit credit to you to have discovered that at an early age. Hey, I'm only mentioning your age because uh oh Michael Finoccario brought it up. He said you started a CAD company before you were actually able to celebrate it. And I was thinking, I was thinking, oh well, he is young, man. But what's your what's the drinking age? Um I'm used to all this partying crowd at University of Texas, right?

Parth 6:56

Yeah, yeah. I think the drinking age is 21, I'm 19. And yeah.

Roopinder 7:02

Okay, all right. So that's yeah, so uh two surprises there. The other one was there's anything at UFT besides football, right?

Parth 7:14

Yeah, yeah. I mean, that's the big thing. Like if you go to the games, it's like it's crazy, it's huge. It's huge, right? Don't they have a stadium?

Roopinder 7:22

Like, how big is that stadium? 100,000 spectators?

Parth 7:25

Yeah, 100,000 people, yeah.

Roopinder 7:27

Oh my god, that's almost like a religion over there, football, isn't it? It is. You're not a you're not a football player, you're a squash player, right?

Parth 7:35

I am a squash player, yeah. Yeah, we actually we actually went to nationals this weekend. We were in Philadelphia this weekend playing nationals.

Building The Beta And Traction

Roopinder 7:42

Oh, you're in Philadelphia. That's my that was my home uh was my home city for a while. Yeah. Yeah, okay, wow. Interesting. So good, how'd you do?

Parth 7:53

We finished second in our division, so it was pretty good. We beat a couple of universities, umd, um, we beat Rice, and um we lost to Northwestern in the end. So that was kind of a bummer.

Roopinder 8:06

Oh, I don't feel bad, you know. I always I always feel bad for number two because they always think they they lose, but I'm thinking, hey, you're number two, right? That's the place. I felt so bad for uh oh, who do the Super Bowl losers? This I was again, I'm in that habit. You know, if Seattle won the Super Bowl, but New England was like, yeah, hey, that's number two. Don't don't be crying, right? Yeah, yeah, yeah. You did pretty good. You should have a parade too. Maybe a smaller parade, right? But have a parade. Well, congratulations. That's great.

Parth 8:37

Thank you. Thank you.

Roopinder 8:39

Yeah, and you're are you captain?

Parth 8:42

I am not the captain. Um, one of my friends is the captain, yeah. Oh, okay, all right.

Roopinder 8:46

Yeah, so I gotta ask about your upgrade because obviously we're both Indian, right? Yeah, what do you your parents say probably are very proud of you being an engineer because that's like that's like number one job for for sons, right? Um what do they say about you playing squash? Anything?

Naming And Vision

Parth 9:03

No, yeah, they're super supportive of everything I do. Um I grew up you should be taking up cricket though, right? I did. So I grew up in India. I grew up in Mumbai and I played cricket there. Yeah, I was playing every single day. I was captain of my school team. It was like my main sport growing up. And then and then we moved to Canada. I I took up baseball there because cricket wasn't big, and the most similar is baseball. Yeah, so I took up baseball there, and when I moved to Houston, I started playing squash. It was like a whole bunch of sports that I grew up playing.

Roopinder 9:36

Oh, okay, okay. Oh, so you were you said you were in Canada. Where in Canada? Vancouver. Oh, Vancouver. Okay, now if you went had gone to Toronto, right? Yeah, there's a lot of cricket there. Huge population of Indians, right? In the Brampton area. Yeah, Brampton, yeah. Yeah, huge, so many Indians. I think it's more Indians than Canadians, but then I'm sure there's like cricket pitches everywhere everywhere.

Parth 10:00

Probably, yeah. I think they have a cricket team as well now.

Roopinder 10:03

So okay, all right. Okay, so you're how old did you say? 19? 19, yeah. 19, yeah. All right, very good. How so you got some recognition of being a startup and some help with I think Amazon and with UFT, I think.

Parth 10:24

Are you yeah, Amazon, UT Austin? We're part of Google for startups as well. We were part of Microsoft for startups, so we get support from them as well.

Why CAD Still Feels Hard

Roopinder 10:33

Okay, and you've already launched Makistry It's actually a working, something actually works. I couldn't find out how to use it actually. Do I have to be on a waiting list or something?

Parth 10:42

No, so it's online. You can go to Makistry.ai and try it out. Um, yeah, yeah, it's live. I think there's a try button on our website too that you can just click on. The try button. Okay, all right.

Roopinder 10:53

Yeah, yeah. All right. So it's all cloud-based, I guess.

Parth 10:56

And uh yeah, yeah, it's all cloud-based, all a web app. We don't want to kind of restrict ourselves to building on desktop apps. We kind of wanted it to be open to everyone, open for everyone to use. Um so we we built a web app. I can show you a quick demo if if you want.

Roopinder 11:13

Yeah, please, please. Yeah. I I think people would rather see a demo than my face. I think they'd mind looking at your face, but probably the demo would be good. Go ahead. Cool.

Parth 11:26

Can you see this? Uh-huh. Yeah. So I just took a quick screen recording this morning. I can just talk over it. Okay. Yeah. So our whole idea behind Makistry was to make it as easy and intuitive as possible. The most obvious way to make was to make it text-based. So it's all text-based, you can just talk to it and um create CAD models, continue iterating on to signs, etc. So when you go to Makistry, this is what it looks like. And you can just put in a prompt. Here I'm making a VESA mount adapter, um, which basically connects a VESA monitor to a mount. Um, and I give it some specifications that I wanted and just click enter.

Roopinder 12:07

So what's in what's the LLM behind this? What are you using?

Parth 12:10

So we're we use mostly Anthropic models, but we kind of specialize them through some AI techniques that we have, and specialize them to work with design and engineering and CAD and kind of have that intuition to work with engineering and design kind of use cases. Okay. And something that's different about Makistry is that like most AI CAD tools that you probably see out there are just text to CAD. But what we want to do differently was to have a middle step that kind of helps you break down that design, flesh it out a little bit more, break it down into features, functionalities, components, um, so that the user kind of has more of a direct connect to the design that's gonna come out. Because right now, like with text to CAD tools, it's it's kind of like a black box. Like if you don't put in the right wording, then what's gonna come out isn't exactly what you had in mind. So we kind of added this step to have an intermediate kind of intermediary between your idea and the design that's gonna come out.

Student Life And Background

Roopinder 13:14

So it kind of tell me how you would control the uh, you know what they call hallucinations of LLMs. They uh, you know, you can ask it one thing and it assumes an entirely different thing or starts giving you. Yeah, because it's trained on you know, garbage. It's trained on uh I'm gonna make fun of. It's not just anthropic, but all LLMs are trained on the vast amount of data that exists in the real world, including like Reddit posts and misinformation, right? How do you keep it from not confining your answers to something that makes sense for engineers or designers?

Parth 13:49

Yeah, so those are mostly done through something called system prompts that you give your LLM. Um, you kind of give it's kind of like a baby, right? You kind of just push it in the direction that you want it to go and kind of give it some guidelines, give it some kind of ideas or like preliminary things for it to think about when it takes the input and gives you the output. Like you can give it some reasoning steps in between that it can think about and kind of specialize it to whatever you need, like the use case that you're trying to give it for. Yeah. So that's kind of what we did for the hallucination part. And then for obviously the design aspect, we had to be more specialized.

Roopinder 14:27

Is that what they're calling prompt engineering? Like yeah, that is that is prompt engineering. Yeah. So this is is this more or less an interface over entropic, or have you done more to it than that?

Parth 14:38

No, so there's a lot more to that. There's like stuff that we're done around RAG with kind of fine-tuned the models from them to specialize them for design and engineering, like I was talking about. So there's like stuff that there's like layers to it. The base layer would be the kind of entropic model that we have, and we kind of had to build on top of that to be able to get it to provide accurate and reliable results. Because the problem that you would face if you just tried to use the plane model, it would just hallucinate a lot, like you said, and it wouldn't be something that an engineer would use. Um, because like you obviously need accuracy, you need reliability, you need consistency.

Roopinder 15:17

Um, so like there's we have to build some layers on top of that to be able to talk about those layers a little bit more because that's important because I think people would might assume hey, they can start using, they they first assume you can use LLMs to make CAD, and then they're then right away they're disappointed that hey, you can't use LLMs for CAD, right? But you've added some layers to make it more acceptable to engineers and designers, right? So talk about each of those layers a little bit more.

Parth 15:50

Yeah, I mean, I want to stay away from going too in-depth with that because I get kind of um that's your AI, that's your uh yeah, that's like the moat that we're kind of yeah, yeah.

Roopinder 16:03

All right, that's your secret sauce. So yeah, okay. So don't yeah, tell me, so tell me in vague generalities or anything, like what they are. Yeah. To be specific. Don't give away the code. But tell me, yeah, make it sound like why is it important what you did? And why can't why shouldn't I just be assuming that it's a wrapper over LLMs?

Access And Cloud App Basics

Live Demo Setup

Parth 16:25

Yeah, a general purpose LLM, like if you just used say sonnet four or something like that from Enthropic, it's obviously trained on all the data that's out there today. Um, that could be everything from correct data to wrong data and the whole spectrum between that as well. Um the problem with hallucination is that it could give you, it could fetch some wrong data and then give that to you. Um, and what you want to avoid is getting that wrong data. You want to get it to get the right data, kind of think about and reason through the things that you wanted to think about. So if I'm an engineer, I would think about these things when I'm starting a design, right? I would think about what the features of that design have to be, what the functionalities have to be, what the geometries have to be before I actually get into the design phase. So um these LLMs are called reasoning models. So they have these layers built into them that help them to go sequentially through and think about the things that you want them to think about, which is the prompt engineering part that I was talking about. Um, and then on top of that, um, there's things that you can do around RAG, which is retrieval augmented generation, which is basically there's specific like documents or specific um param like guidelines or um standards that you want it to follow. And if that's written down somewhere, then it can kind of go and fetch that and think about those specific guidelines or specific um steps or information that you want it to kind of have when it's reasoning through the problem that you gave it or the prompt that you gave it. Um so that it can kind of use that information to aid its thinking process and its reasoning process and then comes out with something that that's usable for engineers. And then there's obviously fine-tuning that you can do.

Roopinder 18:11

When you say RAG, what is that?

Parth 18:13

Yeah.

Roopinder 18:14

Tell me what that.

Parth 18:15

So, for example, let's say I have a hundred-page PDF document that is created by ASME about some specific standard about, I don't know, like screw holes or something like that. And I want my AI model, like here it says M4 screws or mounting holes. But how does the model know what M4 is? Um, so that what it would do is kind of go to that ASME guide, look at what M4 means, get all the information that it needs about it, and then kind of aid its reasoning process based on that to kind of inform the model that this is what M4 actually means. Um, this is obviously a really small example, but this can be used at a larger level to aid all the information and all the reasoning that the model's doing to specialize it for engineering and design and um CAD modeling and stuff like that. Okay, good.

Roopinder 19:08

Yeah, that's good.

Text First, Then Structured Brainstorm

Parth 19:10

Okay. Cool. So I can continue on with the demo. Yeah. So yeah, here. So it helps you brainstorm through the idea that you gave it. Um and obviously you can edit any part of the brainstorm with text prompt. So here I said I want to change all m4 holes to m6. And what it does is kind of goes through the brainstorm, makes the edits that you requested, and propagates it through all the different things. So here it added m6 and also changed the geometries. And then you can just click a button and it converts that brainstorm and all the things that you defined into the um into a CAD model, a parametric CAD model.

Roopinder 19:52

Okay. And it's every step of the way while it's doing it.

Parth 19:57

Yeah, I mean, those are just like a feature list of Makistry that it's not telling you the steps that it's taking to get there. But it in the background, it is reasoning through everything. It's kind of thinking about the things that you gave it, how it can convert that into a CAD model, and then actually creates it and kind of gives it to a kernel and creates a CAD model that way.

Roopinder 20:16

So it's a CAD model. What do you what format is it in? What uh what engine, geometry engine, did you use?

Parth 20:23

What uh yeah, so yeah, we use open cascade because that's the open source, the gold standard of open source that we have right now. Okay. And um yeah, so it's a B rep, completely B rep kind of representation of the CAD model. Um, to display it as in a web app, you have to convert it into an STL. So this is like a triangulated STL, but um, as I'll show later, you can export step files out of this, which are completely B rep, have all the parameters that you defined and stuff like that.

Roopinder 20:52

Oh but does step store parameters, or does it it's not doesn't store the history, right?

LLM Choice, RAG, And Guardrails

Parth 20:58

Yeah, it doesn't store the feature tree, but you can import this into let's say SolidWorks or Fusion 360 or onShape and then continue kind of adding it to an assembly or creating stuff around it, et cetera.

Roopinder 21:10

Okay, got it.

Parth 21:12

And something that would take you maybe an hour to do, this doesn't less than a minute. So that's kind of and you can see it has the cold patterns that we talked about, the conductings and stuff like that, on which you define the brainstorm. Right.

Roopinder 21:28

And so am I am I reasonably sure? One of the things that turns engineers off about using CAD uh prompt to CAD models is that uh it doesn't give you the same answer twice. It'll generate different answers. So, how do you have you controlled that or is have you eliminated that from happening?

Parth 21:47

Yeah, to a certain extent, yes. That's the biggest kind of challenge we face is consistency. Um, because with AI models, it's like you can specialize it to a certain extent, but it's still a black box. Like you don't, you can't really control what it thinks about. You can control the direction that it thinks about, but not exact steps that it's following. So there's some discrepancy every time you give it the same prompt or you tell it the same idea. I think the design that it comes out with is different each time. The degree of the difference between each trial is what we're trying to kind of minimize, right? If it if I give it the same prompt, it should produce the same result. And that comes down to the brainstorm. That's why we give like an intermediate step with the brainstorm. Because if the brainstorm is similar to the other kind of trial that you did, um the design that comes out will probably be really similar too.

Roopinder 22:40

Right. Um can I be reasonably confident in this example, which you're showing here, that they will be true M4 or M6 uh holes, but they may not be spaced differently. Uh they could be spaced different at different arrangements, and I may have to then play with the arrangements, but they the counter bores and the hole diameters, they should be they should be good, right? Uh-huh.

Parth 23:05

So the if I go back to the brainstorm here, so all of the geometries and stuff like that is defined here.

Roopinder 23:13

So I see. I see. Okay.

Secret Sauce Without The Secrets

Standards, Consistency, And Reliability

Parth 23:16

Yeah. So that's kind of what you can use to define those specific distances or dimensions or geometries that you want it to do. So here, if I say the chamfer has to be 0.7, the chamfer is going to be 0.7. And and we have a measure tool here, um, that button. We recently added this. So you can measure anything that you need. So if I measured the height here, it would be three millimeters, which is the same as we define in the brainstorm. So that's the beauty of parametric CAD, right? If I was creating meshes, then I wouldn't be able to control that because it's going to be different every time. And the triangles that are created in STL meshes are gonna have a different kind of positioning every time. But with parametric CAD, you're able to set those parameters, and the brainstorm kind of helps with that um setting of parameters, and you can kind of define those before you actually create the CAD model, which gives the user kind of more control over the design that comes out afterwards. And yeah. And you can also kind of ask questions and like brainstorm and define or refine your idea after the design's created. So here I just asked it, what's the best way to reduce weight for this um VESA adapter? And it kind of helps you brainstorm, looks at the adapter design that you have and gives you some suggestions of ways that you can do this. Um, so here it said add a pocket to the backside, use lighter material, um, reduce thickness, hollow out the center, stuff like that. So it kind of has that engineering intuition of ways that you could solve an engineering problem. And then you can make any edits to the design also with text prompt. So here as I said, hollow out the center. And it goes through edits the design as well as the brainstorm. So it it kind of propagates the changes between the two so that the brainstorm and the design are always consistent with each other. And if you change anything in the brainstorm, it's gonna propagate to the design and vice versa as well.

Roopinder 25:17

So I'm looking at this and I'm looking at I looked at some of the other examples. Plate with a hole, holes, that's good. Uh I saw a pen pencil holder, I think, on your site, and also copied that, right? So uh are you limited in the amount or amount of geometry or the amount of features that can be made? Or can what I'm saying is I was trying to say is is a part limited in complexity because you're just starting out, or is there or not?

Geometry Kernel And Exports

Parth 26:01

Uh-huh. So, yeah, that's actually a great question. So, one of the things that we are working on right now is kind of expan ding the degree of complexity that you can kind of get out of Mega Street. . Yeah. So right now, the only kind of blocker to the complexity is the amount that you can describe with your prompt. So if you can kind of make your prompt really um descriptive and kind of describe everything that you want, it's going to convert that into the design. Um, but there's obviously restrictions to that because there's some things that you just can't put into words, like some things you can't describe with words. So that's why we're working on a 2D to 3D kind of feature within Makistry where you can upload images, sketches, drawings, engineering drawings, and kind of give that as context to convert into CAD models. So that'll kind of expand the possibilities of the types of designs that can come out of Make History. And we're also working on um making assemblies better. So right now it's really good with individual components like this mount. But if I wanted to create like multiple parts together that are assembled correctly, have the correct constraints, they have the mates that you would need to kind of assemble a large object together, a large project in one go. Um, Makistry history kind of struggles with that right now because obviously AI doesn't have eyes, so it cannot really it doesn't have the visual sense of where different parts connect, where um where the mates have to be, where the constraints have to be. So we're kind of working on that aspect as well, where um you can kind of create multiple parts in Makistry and then assemble them on this platform as well.

Measuring, Parameters, And Control

Roopinder 27:59

I see, I see. Okay, so assemblies we can look forward to. So this is all very interesting, and I'm glad to see it happening. The natural language interface to CAD is has been wanting for so long now. Uh I hate the fact that I have to learn a CAD language every time I learn a new CAD program, right? Yeah, I don't want to be a SOLIDWORKS engineer, I want to be an engineer or any of those. I don't want to be an AutoCAD artist. I want to be, yeah, I want to be an architect, right? So any of those things. But I all the time I had to learn the CAD program, which is very difficult to most people. How would you think about what would you think about a future in which I don't have to learn any CAD commands that you know I could just tell any CAD program, not just not just yours, but yeah, you know, I can just talk to it in a natural language and tell it what I want to do. And and AI figures out what the actual commands are. That because as a concept, it seems easy for me. Forgive me, I'm not a programmer, uh developer, right? But to me, it seems easy. Like you AI, hey, AI, you can read all those help manuals, right? All the documentation, you know what all the commands are, so you know the CAD language, right? You know my language, English. You can certainly do translations. I know that's true. So, how the hell can you not learn English to CAD? But that seems to be like a lot less words. CAD has what, yeah, 400 words. So how can you not do that? And then, okay, go ahead. Give me your thoughts on that. Is it very futuristic sort of thing?

Editing By Chat And Design Intuition

Complexity Limits And 2D-To-3D

Parth 29:44

Yeah, this is what I feel like this is what text to CAD does right now, right? It it kind of takes you from your ink your idea in plain English and then converts that into a CAD model that you can use. Right. But one thing that I am seeing is kind of the co-pilot kind of thing, um, where you're in SolidWorks, you see the interface, there's like a million buttons everywhere. Yeah, no one really knows what they do. And then you have kind of an AI chatbot on the side that you can talk to and say what you want to do with it. That was that is something that a lot of kind of companies are doing, but we wanted to stay away from that as much as possible because um, for me, I think the co-pilot kind of sense was really restrictive in terms of the types of things that we could do with it. Like I would be restricted to the whole host platforms interface, the APIs, the design experience, um, and kind of limit the amount of innovation that I could do with it in terms of features that I could add, the type of experience, the design experience that I could give to the user, right? So that's something that a lot of companies are doing, but we're kind of taking the contrarian view on that and going the other route of building something from scratch. Like our mission or vision with what this can move into is it's something, it's like an AI native design tool where you come in with an idea. It helps you brainstorm that idea like we do. It helps you create the CAD model. Um, you can continue editing, creating on top of that. Then with just some words, um, you can kind of run simulations on it, do analysis on that design that you made, um, do CAM on it. Like if you would give into a CNC machine, you can create toolpads or you can kind of slice it for a 3D printer or something like that. And then you can come out with a physical tangible part that you can, or a prototype, or a part that you can use in your real life, right? So that's kind of the idea, like moving away from that manual interface and going towards an AI native conversational, completely text-based, or maybe augmented by some manual insights or manual kind of like the human innovation and ideation is still there, but the execution is done by the AI. I think that's where that's where we're headed, and I think that's where I think that's the future of CAD that we're betting on.

Assemblies And Constraints Roadmap

Roopinder 32:04

You mentioned Open Cascade, which is a uh a solid modeling engine, and uh, and uh I'm thinking of all the things I'm thinking of that not just Open Cascade, but any geometric model modeler as sort of a let's see, what's the right word? Sort of a prison. Sort of a prison. It's like it I am imprisoned to make prismatic shapes, right? And if you look around the world, like nature, it's anything but prismatic shapes, right? So I want I want some help in making designs that are good. If they are truly creative, why can't they be more naturally shaped or curvaceous, right? Solidworks is like horrible at making curvy shapes, right? I have to be a freaking master of SOLIDWORKS to make it do surfaces, right? And yeah, that kind of time. So I avoid it, right? But then I think about oh, if I'm gonna make any kind of organic shape, for example, take our faces, right? I want to take our faces and use them to subtract my face from a mask that I'm making. So it's a it's a it has a perfect fit to my face. And I think, wow, or anything, steering wheel for a car. If I want to make custom steering wheels for cars that are exactly the shape of a race car driver's hands, right? I can 3D print them, but I'm gonna have a very difficult time making those organic shapes, right? And right engines can't do that.

Parth 33:38

So in your in your attempt to I'm gonna say recreate CAD, which would you promise to make these shapes a little bit easier to make Yeah, I think that can be done, it'll just be I can see how that can be done actually. Like right now, if everything is kind of done in a script-based kind of method, right? Open cascade, how you create models is kind of running through a script. Um, if you can recreate that script in a way where it works better with curvaceous shapes or organic shapes um rather than prismatic shapes, like the primitives that you have in that kind of scripting language are are the more organic shapes, and you kind of build on top of those and connect those two and then run kind of up cat operations on top of those shapes. Um I don't see why that can't be done.

Beyond Copilots To AI-Native CAD

Roopinder 34:36

Okay, it should be, right? Okay, but I would say don't make the mistake that uh do you know what topology optimization is and uh generative design? Right, yeah. Don't make that mistake. Don't make these super blobby, useless shapes that you can't possibly manufacture, right? Exactly. You know, the CAD companies have tried to try to uh position that as oh, as AI. They said, oh, we had AI, we had this topology optimization like 10 years ago, right? I go, no, no, no. You never made any shapes I liked, right? I never had shapes I can use. I had to 3D print everything anyway. Uh yeah, don't do that. Make it more useful, make it more machinable or castable or you know, manufacturable parts. Like that was just to me, it was just like lunacy. You're making these parts that have you seen you must have seen them. Some of those shapes, yeah.

Parth 35:29

Those are just not machinable, right? Yeah, you can't do that.

Roopinder 35:32

All right, well, this is great. Well, uh, you said uh you said you use the pronoun we several times. Is there more to McHistree than yourself? Are you a team of people helping you?

Parth 35:44

Yeah, so it's primarily me. We have a team of four or five engineers who are kind of working with me. Um they're also students at UT, so they're kind of but they're really talented, so it it works out. Oh yeah, yeah, yeah.

Roopinder 36:00

How come you can't use that massive brain power that's in India? You can't you can't reach across continents and get those, you know, there's millions of IT people in India. But I and you know where there's a lot of them that are doing CAD work in Pune, right? Right? Yeah, right near Mumbai. Yeah, yeah, yeah. Honestly, every every CAD company has an office in Pune. Yeah, does I think DASA Systems does. I don't know if Autodesk still does, but yeah. But yeah, they're all over there. You'd have no trouble.

Organic Shapes Without Unmakeable Blobs

Parth 36:32

It works. Yeah, yeah. I mean, I could, yeah. But for me, it's about collaborating and brainstorming together and kind of creating new solutions. So having them face to face is really helpful.

Roopinder 36:44

Yeah, that's uh okay, yeah. And you can all meet together and discuss that. Is helpful. Uh I'm only I'm only joking with you. All right, all right. So are you in a dorm situation?

Parth 36:53

Yeah, I'm I'm student housing, so it's it's off campus, but it's like a two-minute walk, so it's not too bad.

Roopinder 36:59

Yeah, where did your parents move to, or did they not move?

Parth 37:02

No, they did. They're in Houston now. Um, and I'm in Austin, so it's it's like a two and a half hour bus drive. Oh, okay.

Roopinder 37:08

That was the scene of the last uh SOLIDWORKS conference. I don't know if you managed to go. Have you ever been to one of those?

Parth 37:15

I couldn't go there um because I had exams and stuff that week, so I couldn't go there.

Roopinder 37:22

That's pesky exams. So with all this, I you I hope you're keeping up your marks, as they say.

Parth 37:30

Yeah, yeah. Still still have a 4.0, so clinging on to that as as much as possible. Yeah, no way.

Roopinder 37:37

Really? Okay. Wow, okay. That's that's very impressive with all this work you're doing on this startup. And you can still add play squash, Ed?

Parth 37:46

Yeah, yeah, it's fun, yeah.

Roopinder 37:49

You don't look like you're losing a lot of sleep.

Parth 37:51

So are you where are you doing all this time? I think it's just about like priorities. I had to give a lot of stuff up as well. Like I I used to make music too. I don't know if you can see my guitar back there, but I had to kind of dim that down a little bit, focus on the things that I want to focus on, and things that are important right now are like make a street school and squash. So those are the three that I chose and that just in that order, yeah. Yeah.

Team, Support, And Collaboration

Roopinder 38:20

All right. But what is your what you hope for? You know, I like to say in the in football, in college football, like you could get drafted as a

Parth 38:30

as a I think you can get drafted as a freshman actually into the pros, right? Don't let that as a rookie, don't let that happen to you. Like don't go, I guess you could do it. Bill Bates did it, didn't finish his college program because yeah, I mean Indian parents don't uh Microsoft, right? Yeah, yeah. So do your do a favor to your parents, get that degree. Yeah, yeah, no, I don't plan to drop out anytime soon. I think yeah.

Roopinder 39:02

All right.

Parth 39:02

As long as I can manage the two, then I can get through it.

Roopinder 39:06

Manage the three.

Parth 39:07

Oh yeah, the three.

Roopinder 39:10

Is there any so how what do they think of squash players at UT? Do they even do they even know that they exist? No.

Parth 39:18

We're not we're not super popular here. We went to nationals and no one really knows about it. Um yeah, we're trying to trying to grow it out. So if there's any sponsors that into this who play squash and are interested, then we'd love to. And if Makistry takes off, then Makistry can sponsor it. We'll see.

Balancing Startup, School, And Sport

Roopinder 39:36

Maybe, maybe that'd be fair. You could say you're a startup, but probably you can't say I'm on the squash team and get any dates. Oh yeah. All right, I'm not trying to pry, but it's just interesting to me because it's like I said in the beginning, that is an amazing football culture. I'm not, no, I follow the pros, but I don't follow college football. But it's amazing to me how these southern towns, southern cities and states it's so popular there. Wow. Okay, I better not keep you, probably have a lot. You have at least many other things to do than talk to old journalists. But I want to wish you the best of luck. I think what you're doing is phenomenal uh and sorely needed. Uh, like I said, CAD, CAD needs an upgrade really, really bad. It just needs to be made more useful. I say it's finally we're able to put the computer aided into CAD, right? I mean that computer aided that was there before, that's a farce, right? It was never the computer-aided design. It's computer-aided documentation, if anything, right? All it could do is make the shape so we we had to think about these shapes in our heads before we put them into CAD, right? Yeah. Now we're finally gonna get there, right?

Parth 40:58

Yeah, that's kind of what we're doing, right? Making CAD easier, faster, and smarter. Um and kind of lowering the barrier so that anyone can create models and drive innovation, but then raising the barrier with parametric CAD so that professionals can kind of drive the next wave of innovation forward as well. Very good.

Closing Thoughts And Next Steps

Roopinder 41:22

And you can then refer to it. You can use it in your marketing. I don't know if you've discovered marketing yet, but a lot of times that's more important than the distribution is key, right? Yeah, yeah, right. So you can uh refer to lines from it or whatever comments. So again, best of luck to you. Great meeting you, and I hope to you too. Maybe we'll let's promise to meet sometime in the future, and I will definitely have more good news. Yep. All righty. All right, thank you. Great to meet you. Have a good rest of the day. You too, thank you.