Owein Dourneau, CEO of MecAgent, Converts Natural Language to SolidWorks

We dive into why CAD feels hard and how natural-language automation can remove friction without forcing a platform switch. Co-founder and CEO Owen Dourneau explains MecAgent’s approach to compiling plain English into SolidWorks actions, the limits of file translation, and where AI can truly help engineers.

• Pain points with feature trees and steep learning curves
• Why building on top of SolidWorks beats starting from scratch
• LLMs as a CAD compiler for dependable automation
• Examples like DXF export and sheet metal unfolding
• Simplicity over code editors to reach non‑programmers
• Risks from incumbents and startup speed advantage
• Drawings automation tradeoffs and scope
• Library, remixing, and sharing of automations
• Roadmap toward more autonomous CAD agents
• Auxiliary tools for engineering Q&A and part finding
• Partner discussions and realistic adoption paths

Transcript

Roopinder 0:21

Hello and welcome to FoDES, the Future of Design and Engineering Software Podcast. My name is Roopinder Tara. On the show, we will have guests that will discuss tools and technology that engineers will find interesting and useful. Hi, nice to meet you as well. Thank you for joining.

Owein 0:38

How do we actually pronounce your name, Roopinder? That's great. Where is it from? Originally from India. Nice.

Roopinder 0:43

You're originally from France?

Owein 0:45

Yeah. Did you guess with the accent?

Roopinder 0:47

The accent gives you a way for sure. But you are living in San Francisco now, right?

Owein 0:52

We are partly in France and partly in San Francisco. MecAgent is actually an American company, but we also have a subsidiary in France, and we are using both worlds. I would say Europe and the US. But yeah, I'm the one in the US.

From France To SF And Setup

Roopinder 1:05

Okay. And your offices are in uh Fort Mason. Is that correct?

Speaker 1:09

We are currently in Fort Mason. I don't know if you heard about it, but very cool place. A lot of cool people there.

Roopinder 1:15

Yeah, I've been there uh a few times. Um just north of you up in Marin County. Have you had a chance to visit yet?

Speaker 1:21

I wouldn't I've never been there, but I would love to visit the coast and around San Francisco and all that.

Roopinder 1:26

When we first moved to the Bay Area, when I showed my county, she was she finally overcame her reluctance to move here and said, This is nice. It's not what she had thought of as the Bay Area, but Marin County is the jewel. I probably shouldn't be telling you this because Marin County doesn't want anyone else to move there. Close the door behind them, it seems. But yeah, you just cross the Golden Gate Bridge and it's like a whole new world.

Speaker 1:49

I've kept myself from that to not be too distracted for now, but I will definitely check it out.

Roopinder 1:53

I'm sure you're too busy working. I want to say your name correctly. How would you say it?

Speaker 1:58

It's just Owen. Guy is just there to mix up.

Roopinder 2:01

Just like Americans would say it. Okay, that makes it easier.

Speaker 2:03

This way of writing is actually from Welsh, but it's just pronounced normal. You could say Owen, but no one calls me like that. It's just Owen.

Roopinder 2:11

What does your mother say?

Speaker 2:12

Owen.

Roopinder 2:12

Okay, good. Now you've started this company with some co-founders, right?

Why CAD Feels Hard To Use

Speaker 2:15

Yes. I started a company with Alexandre and Guillaume. I've met them during the story we met during studies in France, but I actually met Guillaume during an exchange in Sweden. And we instantly connected. They are both mechanical engineers, and I'm the software learning mechanical engineering on the go as well. I I actually did some mechanical engineering during high school and afterwards, but I promised myself that I would never touch again SolidWorks. And turns out it didn't really work. I actually doing more SolidWorks than I would have ever thought.

Roopinder 2:46

Most mechanical engineers use SolidWorks, but tell me what happened to you. You experienced SolidWorks and you were disenchanted with it?

Owein 2:52

I was more interested at that time in software than mechanical engineering. And so I was already knowing that what I wanted to do was more around AI at the time. So that was already like a long time ago when nothing about Gen AI was out there. But I was already knowing that every time I hear the noise of the error in your feature tree or whatever, I was like, I don't want to hear that every day. SolidWorks is a great software. The math behind, like the whole kernel behind is not even from SolidWorks actually, but the whole CAD is great. I have to admit that it's also a lot of the time very painful to work with those tools. Not only SolidWorks, but generally CAD, I would say. I've never met so many people complaining about the tool they use every day.

Vision: AI That Truly Aids Design

Roopinder 3:31

I think I've used a lot of CAD packages, maybe a dozen. I've mastered a few, but I'm of the firm opinion that CAD is very difficult to use. Unless you're a user, then it's very easy to use. But for everybody else, people that are trying to learn it, people using it infrequently, it's extremely difficult. It's actually a barrier to design. It's not computer-aided design at all. So what I'm hoping for, people like you, right, to change all that, to make computers finally helping us with design.

LLMs Controlling SolidWorks

Owein 4:01

That's the initial view of what AI should do for mechanical engineers. Help the already experienced mechanical engineers and make everyone able to be a mechanical engineer. The CatSoftware is a tool, but at the end of the day, the mechanical engineers have to think about like the engineering part, like the design part is a thing, but the engineering part is also more than just the design. To me, it's a tool that should be easy to use, which is not the case today. The learning curve is very steep. AI is definitely something that can get you there, help you if you are able to ask any question and then get instant feedback. And you should do like this, you could also do like this. And I think that's the same. Like the playbook is similar to what is happening to software development right now. And that's the initial that's the initial thought of like why MecAgent. It's like we've seen how AI revolutionated the way we actually do software today, right? What would have taken like a month would take like a few weeks now or even hours just because it is now way easier. There is a bunch of phonetics coming with it. I would be happy to talk about it because it will for sure transpose to like mechanical engineering, which is a lot of code is bad. A lot of code is like over like they some people like sometimes it's overused. So you have this like technical technological depth that you're building up. But at the end of the day, I would have never been as confident as today to say that my grandma could actually do a B2B app and ship it like very fast because of all the AI tool out there.

Roopinder 5:20

Most people know AI. Let's just say civilians know AI for because of ChatGPT and others like it now that have used LLMs. But you found a way to use LLMs to control SolidWorks in this case, right? So would I be able to use a natural language interface, like type out a command, and MecAgent agent then can actually turn that command into SolidWorks commands?

Automation Over Co‑Piloting Today

Speaker 5:43

You could do that. We have what we call the CAD compiler in the software, where it's like the agent is directly taking actions in the software. It works well when you want it to be like reflective about oh, let's say I want to change all the names of my parts into another language, like all the names, or name them correctly. Then you have to have this LLM reflecting of on okay, what's the part look like? And okay, what language is it right now? What language I need to add it. So that's one part which is very experimental, not definitely not the most useful for now. The other part that we like pushing today is creating automation. This is exactly when you were talking about the steep learning curve earlier. There is something that even the most experienced mechanical engineers are not using in their CAD software today. It's a scripting language. But when you think about it, it's the most powerful tool. Let's take the example of SolidWorks. Everything you can do in the UI, you can also script it. Today, the goal of this feature for us is to make everyone able to do automation in their CAD software. So you would go, hey, I want to export all those parts, all those sheet metal parts of my assembly to stuff, but they need to be unfolded and all of that. And so you will go step by step, typing out maybe the automation is not working first time, then you refine it. We were talking about making it easier. We definitely pour like import every mechanical engineer now to be able to create automation in their CAD software.

Build On Top Of Existing CAD

Roopinder 6:57

Let me get an idea of your general philosophy. So you came into engineering, you discovered that a CAD is not really helping me, it's very difficult to learn. There should be a better way. But you chose instead of making a whole new CAD program to work with the leading CAD program, right? Tell me your thinking behind that. Why wouldn't you just say, I don't like the way SOLIDWORKS works? I'm going to make a new SOLIDWORK.

Speaker 7:21

The first thing I thought was, let's make a new SolidWorks. I talked to Alexandre and Guillaume, and they were like, that would be very fun as an experience, but it's actually not realistic. The real problem is how do you get AI with a mechanical engineer? There are those people that call like on shape, right? 12 years ago, or they came out with this idea of, oh, actually, file versioning is very bad in the CAD world. We're gonna create a full new CAD software in the cloud that can like help you do file versioning, which is a very good, like this is a huge problem. They were solving a real huge problem. Still 12 years after they're getting bigger and this is going in the right direction, but they're still owning a few person of the whole market. So for us, it was the same idea of how do we be, how are we the fastest to be associated to AI for mechanical engineer? The answer to that for us was we have to be the ones that are actually integrated into real workflow. The biggest problem right now is that if you're using SolidWorks, you can't go to Fusion 360. You can't move from one software to another. We understood that and we're like, okay, let's just be on top of all the softwares to learn from the mechanical engineers use because Alexandre and Guillaume are mechanical engineers. You can't assume you understand all the needs, and you can't assume people will tell you what they need. So we were like, let's create a compiler. We solve existing problems, such as like creating automation for [...] use task, but also part finding, also answering engineering questions and all of that, which is already an added value. But like we, and for the most important part, learn about okay, how do they use it? Oh, interesting. And then maybe that will be one day, and I think AI could be the biggest leverage of how how do we change the cat world right now? How do we actually change this whole for Oligo Ball, right? For company of all. And I think AI is the leverage. But for us, the strategy was gather the data in terms of use cases, how do people actually what do the people actually need? And maybe one day we'll end up being doing a new CAD software, but it's definitely not other route. But there are some people who have creating a new CAT software based on their own kernel, which is very cool. But realistically, if you want to move like the fastest and go in the direction of being used by the people, for us it was a no-brainer of okay, let's integrate into their existing workflow, otherwise they won't probably because even if the willingness to move was there, making like an entire company move from a CAD software to another is like a very hard task.

Market Strategy And Lock‑In

Roopinder 9:31

That was a smart move when you look at it in the respect of use the most popular CAD program and just make it work better rather than convince people that your software is the greatest and latest. You mentioned Onshape. Onshape had a similar problem. Even though they were created by the founders of SolidWorks, they did not overtake SolidWorks, they had a problem being accepted. So you've overcome that barrier, the most popular mechanical CAD software.

Speaker 9:59

The Onshape move is also good for the market as well. But yeah, that's definitely the problem of it. And maybe in 10 years there will be a majority of people using Onshape, maybe because schools are using it, people learn because it's free and they have a good strategy on that. But AI is not waiting 10 years. The biggest problem right now is translating the file. If tomorrow a company was coming with like a new way of being like, hey, you can't go from an SLD PRT to whatever CAD software you want, reconstructing the entire feature tree without any loss, then the whole CAD market changed, like the whole dynamic changed, but it's not yet there, and I don't know if it will ever be there.

File Translation Limits And Reality

Roopinder 10:36

A common format that preserves all the history, the parameters, and all the design intent. That's just not there. You work with industry standard files and you lose all that information. So yeah, there's not a common format. That's true, that's a very big problem. Your problem takes care of the beginning part at least. You can make a command compiler really for not just SolderWorks, maybe it could work for any CAD program. So any CAD program could have this a natural language interface, like like you're talking about.

Speaker 11:05

Yeah, so we starting by SolidWorks because it's the most advanced in terms of API. But ultimately, the idea is to actually scale that on every CAD software. Every time we're going to introduce a new CAD software, there will be some differences for sure, because if the API is not necessarily as advanced, or the use cases of the automation are probably different, then there will be some differences. But ultimately the goal is to be able to automate any CAD software. We're working so far, we're working with the API, but we're also working on other ways of doing it to be like to be able to scale it to any CAD software.

Risk From Big Vendors And Speed

Roopinder 11:41

Are you looking over your shoulder to see, hey, is SolidWorks understood that this is a their software is hard to use and and they're going to do what you're doing? And they have they have a lot of people they could bring to the table to work on this, right? So is that a thought?

Startup Leverage And Focus Areas

Speaker 11:57

I think it's a thought for every startup funder. Uh the big actors in the space could always overtake you, and especially more for us because we're working like we're integrating with those cat software. So ultimately, since we're sitting on top of them, they have the control. At the end of the day, it's their software. What I think is we have a very good, like our best point is we're a startup, so we can move fast. That's why startups actually exist, right? It's because of those big companies, and when you think about soil works, you think about the soul behind it, right? They're very slow, they don't have the same constraints as us, right? They could be like, hey, this automation thing is actually very powerful, but they can't really try it because it applies so many things, right? And so I am not that scared about and even if I was scared, someone has to try it. Creating startup is like gambling on the fact that you will be faster than the other doing it and better at it. And I think we're going in the right direction because like automation has been out there for scripting in SolidWorks have been out there for 20 years, and that applies to also all the other things, like drawings, for instance. There's a lot of AI around generating drawings. Solidworks and all the Cat Software have been working on it for sure for years, still that there is no clear features or no killer feeders on those software, and there are like plenty of startups that can live with that. So I think it's a risk. That's the first thing we thought, and that also drives the direction we take, like things that are very abuse that like those CAD Software will have it, and we probably won't work on it because they will probably have it. But for the things that are like vertical, as such as like the automation creation and all that, I think we are like trying to be the best at it. So, worst case scenario, it makes no sense for them to start working on it since we are, and so they try to integrate us. Right. We are actually in discussion with the cloud provider, everyone is interested to see what we're doing. It's a thought for sure, but it's the purpose of a startup, right? Doing something that the big company can't because they're too beat.

Roopinder 13:46

I like the way you're heading. Another thing you might watch out for is that these sleeping giants may wake up and realize they are not as fast moving as a startup, and they start acquiring startups because their own developers are not in that mindset. They're not that fast, not working on the right things. They might be using existing code too much, but they would start buying startups to give themselves a leg up, or worst case, to remove them from the market. That has happened. Just today I heard about Cadence. They're not in our industry, they're in electrical design automation, EDA, but they just bought up AI startup. I think probably for exactly what I'm talking about, to give themselves a quick boost in AI that maybe their own people can't handle. How committed are you to staying the course and not being bought out? Or is that an option?

Long‑Term Vision Beyond Macros

Live Demo: DXF Export Automation

Speaker 14:34

Honest, I never thought about creating a company to get bowed by anyone. Of course, I would never say never, but the goal for me. I feel like CAD industry has been there forever and handled by four big companies, five big companies that are no longer really pushing and no longer have incentive to actually make their workflow, like their the CAD itself better because they they don't need to, like they are making so much money out of just being existing and like supporting their CAD software. And my point of view on it is there is a huge leverage to shake that tree, right? Make things differently, maybe not being a threat because that would be known to be a threat, so it makes no sense. At the end of the day, we want to make the mechanical engineer, the end user going faster. So my goal is just to make so it works easier, so it works faster, for instance. But if you look at the long-term vision, AI is in the mouth of everyone. I truly think that not necessarily LLMs, but LLMs gave me the ability to be an expert, not an expert, but like knowing a lot of things on a lot of different topics very fast. So I think even if the end technology is not necessarily like an LLM doing action in your CAD software, but maybe like it will allow that researcher at MIT to actually be faster developing the next generation of models that will overcome the way we do CAD. Maybe, and I hope like in in a few years we'll be doing CAD differently. Or it's not necessarily text because trying to describe the very complex shape of a car by text is definitely not the way. I can see CAD software evolving or new CAD software coming up, like the instructions. We are so set on that feeder tree because it was the way we were doing it for years. But I believe that you could ref re rephrase those instructions. It's no longer necessary to probably have a sketch and then an extrude on that sketch. It could be a cube directly. If you redo that set of instructions, it opens a new way of doing cat. I think AI can be the things. To answer the initial question, I'm not initially creating a startup to be bowed by any companies, but I do create a startup to actually check the tree of the cat world because it has to be shaped. For the sake of the mechanical engineer, I'm not trying to replace mechanical engineers when you're evolving in those cat software that has probably no incentives to make your experience better because they are like owning the whole market, then there must be some startup checking the tree. Show us. I will just start so I'm developing on a MacBook, so it's on virtual machine machine, virtual computer. I don't have a mouse, so I won't be able to move too much around. But here you recognize that it was SolidWorks, right? Right on the left, and on the right is MecAgent. Today MecAgent is like a fully fledged software, so it's not like a plug-in or whatever. It's a software you download from our website and you can install.

Roopinder 17:12

So you're outside of the SolidWorks interface, just a window next to it.

Speaker 17:16

Yeah, I'm just a window next to it. This is how the people using MecAgent today are using it. They put it on the side. Our main feature today is creating automation. So I have a bunch of examples here that I'm doing for the demos. The use cases of those, like the direct use cases of those, is like what we've seen so far. I would say two things. It's because it's important. Like we're talking to the people that were already doing some macro space. There are people in companies paying other companies to do it. And those are the direct like users, the early adopter of this, because they can see the direct like use case, right? They know because they already created automation before. But there is 95% of the rest of the mechanical engineers. For instance, Alex never used any automation because he didn't even know that it was existing. Those people we have to educate a bit on okay, what will be the use cases and all of that?

Roopinder 18:02

I've heard figures as high as 95% don't use the macro features because they don't want to be programmers.

Speaker 18:09

That's fair. You don't have to make it like it's exactly the choice I didn't do actually. I was like, I want to be like someone doing like code, but I could have been the other way around back in high school, being like, I don't want to code, I don't want to see code at all in my life. And so that's the assumption we do. And so, for instance, here there is a good one. So you have this barbecue, which is like a shit metal assembly. I could go for this one, which is a good example. Exportation [...] is [ export, is a problem a lot of people are facing, for instance. And so, like, I could go like create an automation that book exports all the parts in the current assembly to DXF, but all shit metal parts before exporting and create a dialogue. This is also very important, is that we also give the ability to actually create interface on top of so it works. So the idea is that you augment the software, you can create a full new I've posted on LinkedIn recently, a new dashboard we created to monitor certain things in your assembly. So that's the thing you can do that would have taken a long time before. You would have never actually taken the time to do it because if you don't want to learn the code, you don't want to learn the code to actually make on top of it. So as you I just chat and then it creates the automation, and then I can just here I'm just gonna go for this folder. And this UI was generated AI, and then you're just gonna run the automation. And it's gonna go through like all the parts and take the ones that are like metal sheet, and for the one that are like metal sheets, then unfold them and export them to the folder I picked. I could go to the folder to point it.

Roopinder 19:33

I have a question if you don't mind me interrupting. Okay, so your macro was created outside of SolidWorks. How does it know which parts to interact with? Or does it have placeholders where it doesn't know part names, for example?

Context, Safety, And Execution Logic

Speaker 19:46

As you can see, we're just crashed. The question is how does it know how does it get the context right? How does it execute the automation? Or is it smart in executing the automation? Is that the question? Yes, that's the question. So, two answers to that. The first one is right now, the tool itself, when you create the automation, the thinking part is when you create the automation. As soon as the automation is created, there's no context over thinking part anymore. There's no that will say the result of the automation was that, so I have to adapt it. But that's something we're working on. It's another feature that is behind.

Roopinder 20:19

Let me try to be a little more specific. Maybe this will help. You said export all my parts to DXF, right? Now, how is SolidWorks gonna know what this is? This has to be a certain thing in its assembly, right? Which you would point to if you're in the program, but when you're making a macro, you can't do that. You can't point to it or give a name. So how do you overcome that?

Selecting Parts And Smart Tools

Speaker 20:40

The first thing first, it's we have like occasionally it's fully connected to, so it works, right? You can see here that I have the name of the assembly. Okay, so you're accessing the file, you're accessing the model. So we actually do not retrieve any information on the model, like for IP reason. But when you think about that problem, we have all the tools to actually deduce what you're talking about. All the tools that we gave to the agent will help him to deduce. Okay, if he's talking about this assembly, I know because I'm connected to the assembly. When I say all the part in the current assembly, I can have a tool that will list all the parts in the assembly and then execute whatever I ask for all the part in this assembly. It's important for me to emphasize that we're not retrieving any CAD. The only thing that is safe is the chat and the automation that is generated. The whole deducing logic is implied by the tools we are using to give context to the agent.

Roopinder 21:32

Am I at any point able to put in the macro where the program could actually say something like which part? And then the user could point.

Toward Autonomous CAD Agents

Iterating, Favoriting, And Sharing

Speaker 21:40

The way I would do it is probably like generate generate automation that let me select the part and then apply something on there. I would generate an automation that would find all the let's say I want to apply something on holes. I need an automation that like actually finds all the part which has a hole in it, or that let me select all the parts that are that has a hole in it. And then but you could see like the problem two ways. You want to be able to select, but you can also ask the AI to actually do it for you. Obviously, that there are whole in a part, then it's a whole logic that is code at the end. I see. But the next step of that is creating the automation. The direct next step for us is then having an agent, being able to generate those automations, execute them, retrieve whatever the output of the automation, and then modify probably the automation itself and go more on an autonomous behavior. And that's why we like going through with the experimental features, which we call the CAD co-pilot. It's a more autonomous way of giving an instruction, and it will be able to do multiple things. What we observed was being able to generate automation for mechanical engineers and being the best at doing that is something we have to work a lot on. That would be very useful for a lot of companies. So the focus is on it, that's the reason. That this fear seems to be the one that resonated the most. A lot of people were also like, okay, so let's say before it was a good example because the automation actually crashed because of SolidWorks, sometimes happens. But let's say I was like, okay, this automation is exporting in DXF, but I would like the top view of the part exporting in PDF or PNG or whatever, or do anything else, I could go like actually create US that or and then you will go to iterate with the agent on the automation. And as soon as the automation that you created is good enough or fitting one of your needs, or that you want to be able to reuse it, so you will see that it's going through like a whole stuff and stuff, which is quite similar if like oh cursor or whatever the copilot for software engineer works. As soon as it's like good enough, most of the time it will take probably one, two, three minutes to generate the automation. You could add it to your favorites, that will just wait for it to finish. The idea is that you add it to your favorites and then you have all your automation. Our view is that everything that could take two or something to do every time that takes you like two clicks, it's not a lot, right? It's not taking a lot of time. And you would probably before MecAgent never create an automation for that because creating the automation seems to take a long time. But at the end of the day, if you were able to create any new button in SolidWorks to do something you are doing a lot during your cat journey, then it would be worth it. And that's also a bit of the philosophy we have. It's more of like automation. You can do like very complex automation scripts, but it can also be like very simple ones, creating new buttons in your new interface, new dashboards or whatever you you could use to go faster. And that's something you would have never done before because you don't want to learn VBA, you don't want to learn C.

Roopinder 24:30

I think my programming skills stopped at a FORTRAN, and that was like 30 some years ago. So yeah, I don't want to pick up those code again. But here's what I'm thinking about. One of the things that's very distasteful to engineers or things they don't want to do is make drawings. Are you able to automate any of that process?

Speaker 24:47

The drawings is very interesting because it's a huge market as itself. Automating drawings is probably the most obvious use case for AI right now. We have basic automation around drawings, which are not as deep as what a draft aid AI could do, for instance. That's not necessarily our focus right now because it has to be your main focus. It's a whole specialty, it's a whole thing. So my point of view is that CAT software is, and if you look at SolidWorks, there is there are new features around drawings, and it's getting more and more around like the drawings generation part. It's getting more and more like advanced what you can actually do. And most of the time, some people don't even know that there are like some new features in their CAT software that can automate drawings. For us, we have use cases of automation around drawings. You can export multiple views automatically, but it won't be like as advanced as if we were like doing it like full all in. You can automate things around drawings, but it would probably not solve all your problems.

Roopinder 25:44

With drawings, you're actually chasing something that's dying off. What's a better way to barbecue dinosaur burgers when you know you're more chasing the beginning part of the design, not the dying end part of the design, which are drawings, right?

Simple Buttons, Real Time Savings

Speaker 25:57

Yeah, yeah. It's funny because when you think about it, drawings could be God are still that is there and it might be there for the next 20 years, right? Because of like because when you look at it, it's most of the problem in CAD today are because of legacy. And that's the same for the CAD software themselves. Like they are there and still existing because of legacy. Yeah, somebody that is 20 years old won't be able to be moved from any CAD software.

Roopinder 26:22

Tell me this about the automation now. Can I see the automation? What does it look like? Is it lines of programming of code?

Speaker 26:28

Right now, our approach is abstract as much as possible the way the automation works behind as simple as possible, more we have to compare. It's still in discussion of what the direction should be like a code editor for macros. Right now, we're not taking that approach because the idea is to be as simple as possible and to go for the bigger crowd, but we definitely not excluding to be like also providing options to actually get access to the code of the automation. But the thing is also it's gonna change the way like today, it's like only about like scripting and some other secret source, but for other software it will work differently. So we're thinking this the make agents is thought to be like craft CAD, at least for us right now. We're trying to be as agnostic of the software as we can.

Roopinder 27:14

Understand? It reminds me of a quote that's attributed to Einstein. Make it as simple as possible, but no simpler.

Speaker 27:20

For us, yeah, trying not to optimize for this very small crowd of okay, macro creator people, at least for now. Maybe in three weeks, everything will change. That's also the game of the startup.

Drawings: Scope And Tradeoffs

Roopinder 27:30

So, what is the end product? Once I create the automation, which is essentially a SolidWorks macro, where is it? Is it a file? Is it something that's just read into SolidWorks or how does it exist outside of SolidWorks?

Speaker 27:42

So just showing you the modification I has was totally done. So the idea is that as soon as you created your automation, you can save it, and then you will be able to like find it back into your like own saved automation. And so that was the thing I was talking earlier. If you go, we have this like compact view mode. You can see my desk is a mess, but we have these compact view modes where it's like all the automation are created as if they were I see.

Roopinder 28:10

So it's saved inside MecAgent. I like it. It's exactly the way Chat GPT works, it saves all my conversations so I can go back to them. I don't have to redo that.

Speaker 28:18

Yeah, exactly. So I will show you.

Roopinder 28:19

You have all the automation, but you can also go back to can I search in there too? Once it gets to be like days and months of this, inside there for the last command that DXF exports.

Simplicity Over Code Editors

Speaker 28:30

I could search all of them. It's actually a good suggestion for now. I will have it in the pipeline, and next win you can. Thanks. Right now, where we actually implemented, I don't know if it's on this version, but you would have also added a bunch of automation, the good starting points. So you gotta start up with a little library of a original library of the ones that were the most asked by people, basically. And then you can filter it by let's say I want drawings, then it's just the drawings. You have a bunch of what is very powerful is that most of the time the problem with automation was that you had one that was nearly fitting what you needed online, but you would not be able to modify it to make it fit like to your need. And so here it's also like the same with solving the same problem. It's just like you go add to my automation and see this automation that I copied from the public automation part, and then you can remix it. I see. And so that's the whole workflow. But let's say I have my automation here, I want to share it to one member of my team, I can just share it and it will be able to import it and remix it in the same way. Automation for me is like the first step, right? When we created MecAgent, we had a bunch of IDs. Mechanical engineers have a lot of problems. Let's solve all of them. We decided to narrow down a bit more on automation creation right now because it was the one that was like the most used from all the features.

Roopinder 29:45

I think it has the greatest potential to be used because the whole barrier to automation could just be removed. I don't have to learn how to make macros. You can just ask it for automation.

Where Automations Live And Scale

Public Library And Remix Workflow

Speaker 29:55

Another barrier that is going to be removed is like how do you get an AI to talk to a CAD software. And I think automation today is just a way to automate tasks. But when you look, what is behind it, it's a way to use a CAD software precisely. So now we're thinking of, oh wow, it's like actually opening for the mechanical engineers themselves, but like for the entire like AI perspective as well, it's like a way deeper topic. MecAgent started back in March. We were having prototypes of part generation. It's a hot topic in the space. Everyone is talking about like 3D biometrical part generation. We are working on that as well. It's a deep tech research subject, but it's a step forward, like in term of okay, as soon as we were mentioning earlier this like universal file format, and I wish maybe we'll be the one creating it. But as soon as we have this file format, how do we actually get this file format into the already existing CAD software? And it has to be in a certain way. And I think automation and the things behind it can be the way. I was mentioning that because that's the first feature of MecAgent. That's the one we are focusing on right now. But there are like other more cool features to have, and we are working on like being a couple, not necessarily only an automation creator. Today, if you were a user and being like, okay, what can make engineers, what should I know of make agents? I would go for it's an automation creation. There are some other features, such as if you have questions about like certain tip topic, you can go like, hey, how do I compute pressure loss into a pipe or all of that? And then you will get all those PDF, like the answer, but then all the PDF quoted as well to actually verify that the is not getting crazy. And then we're also working on different features such as part finder. We're working close by with catalog providers, and yeah, it is to be able to, hey, I select this axle in my assembly, I want a bearing that fits the constraint. Boom, you have a bunch of bearings. I won't say that you will find the exact one because we're not there yet, but you have a bunch to look through, and it's already making you win sometimes. And this is the whole philosophy of mechanics. What are all the features that can make a mechanical engineer spend more time on engineering more than just like looking through PDF of like constraint or whatever?

Roopinder 32:07

Oh, it has have you become a SolidWorks partner or they have a third-party program. Do they allow would they allow you to become a partner or have you already done that?

Speaker 32:17

It's an ongoing process. We're discussing calls with that so like this whole like they have, as you mentioned, like the pro like the partnership program, and we actually talking with them right now. Good luck with that. It's always cool to talk to people that actually have the right questions.

Roopinder 32:31

I'm a frustrated CAD user, as I said. This is music to my ears, honestly. I really appreciate the conversation. I'm glad to hear you say that you have a lot of the same gripes against CAD as I do, and you're better than me. You're actually doing something about it. I'm just complaining.

Speaker 32:47

Always cool to connect with people that understand the pleasure to talk.

Roopinder 32:51

I hope to get you back on as a guest. Thank you very much. Okay, thank you. For listening to FoDES, the Future of Design and Engineering sSoftware Show, brought to you by ENGtechnica. I hope you have learned of a new application or technology that will help you with your job. If you have an application you think would be of interest to other engineers, please let me know by emailing me at roopinder at engtechnica.com or message me on LinkedIn.