Designing Smarter: AI’s Role in Modern Engineering | Part 2

Raymond Yin:
Welcome back to The Tech Between Us. We continue our discussion with Nemanja Jokanovic, Head of Sales at SnapMagic, exploring how artificial intelligence fits within the design process. Let's jump back in.

So, something like this, once it becomes ubiquitous, once this becomes the standard it's going to change the way we design products from a workflow, from an efficiency standpoint. Change it completely.

Nemanja Jokanovic:
Oh yeah. I mean, it's already happening, and this is why you start seeing people, big companies and some startups, my startup too, really starting to focus us at SnapMagic and launching the copilot. AI enabled copilot, focusing on evolution of component centric development to reference design, electronic development, and even further maybe subsystem electronic development. And I don't know if you are hearing this, but a lot of conversations that I've been a part of walking through Electronica, walking through Embedded World, everything, a lot of conversations are evolving onto that subsystem level of electronic development.

Raymond Yin:
Yeah, I mean, I've talked to various companies. Obviously, you guys over at SnapMagic, talked to other people and that really, there is a very common thread running through when we talk about, okay, what's next for design tools?

Nemanja Jokanovic:
And everything we discussed so far, I think leads naturally leads into more. Right now, I don't have to spend so much time on focusing on one thing now. I have incredible capabilities of artificial intelligence and just natural languages that I could use for the development of my electronics. 

Raymond Yin:
So how do you think that could, I mean, I know you guys at Snap are working on integrating a lot of this. I mean, how do you see it working for an engineer in a design process?

Nemanja Jokanovic:
Yeah, so you just think about where does electronic development actually start? And electronic development – a lot of people assume that electronic development starts in the design tool itself, the CAD tool. Oh, I'm just going to start designing things and dropping components in. The development actually does not happen there. The development of a subsystem or even just a single board starts happening in an engineer's head. The problem with that is that today a lot of people are still sketching ideation process of electronics and research behind what they need - still happens on a component manufacturer side. Maybe I'm trying to design something that conceptually I think it's going to work, but most of the design starts happening from something that already exists. And that research just the sheer volume of research that happens for you to even get to the point of laying out your board takes a tremendous amount of time.

Raymond Yin:
Absolutely.

Nemanja Jokanovic:
Today, majority of people are still designing electronics or sketching the ideas. You start your search in Google, then maybe you go to component manufacturer site, then maybe you download a couple of reference designs. It's like, okay, good starting point. Then you start looking at components that kind of go around centralized idea that you're trying to make, then you want to see how that board itself would work with a separate portion of the subsystem. And that takes time and it's still being used in Word, in PowerPoint, in Figma. But the problem with that Ray, is then you move on, you do your job, and you move on and come back well, you lost all that research unless you're really storing it in some kind of a folder and you're an extremely organized person, which a lot of engineers are. You lose a little bit of a context of what you were really working on. So, you can, start right now, your board in the CAD tool itself. And a couple of things that we are trying to solve is - as you're searching for components, AI can just help you decide what goes with that. Now, components that are relevant to maybe you're trying to find a specific MCU that's going to work with low power, and needs to deliver a certain volume of output, needs to work within certain temperature range, and all of those things.

Raymond Yin:
Right? Certain level of performance. 

Nemanja Jokanovic:
You need to connect it to a specific USB and USB connector and just going to start thinking about those things. And as you really build the core concept of what you're trying to develop, then really the application of AI today can be … well, populate the passive components as the result of this input or add additional components that could work well within the constraints of just what I'm trying to achieve with my design. And the AI itself is as good today as you want it to be, which is important. And one thing that we particularly focus on is really the quality,

Raymond Yin:
Which is absolutely critical, obviously in designing products.

Nemanja Jokanovic:
I mean, you can't just use GPT, especially for such a high standard industry such as our industry. I mean, things cannot fail at the end of the day,

Raymond Yin:
Right? It's got to work.

Nemanja Jokanovic:
Exactly. So, I think this is why we are a little bit slower to adapt the application of artificial intelligence. But in our particular instance, what makes me excited is that what goes into the AI, it's not just a natural conversational language. It's really about training the model. And the way we do this is, we work with pretty much any component manufacturer you can imagine on standard reference designs and making such reference designs digitized and interactive. And then you can really start training the model around already pre-approved reference designs at a much larger scale. This is where really the electronic design becomes fun.

Raymond Yin:
Especially over the years, and especially more recently, manufacturers are not releasing components unless there is some sort of reference design to go along with them either in a form of a board or a reference circuit. And so, you guys are taking that information and training an AI model be able to create designs based on those.

Nemanja Jokanovic:
Yes, yes. The short answer, yes, it's really about not just using the language. Anybody can just apply a GPT, but when you apply math and when you apply user interaction and volume of interaction on top of manufacturer reference designs combined, that's an extremely powerful solution. So, training the model around those specific things is very powerful because then it can come - not just serve in a context of reference design of its own, but it can serve an engineer in a context of their own application, which is maybe I really care only about these things, maybe temperature, and I need to connect to a specific … let's say I'm designing a drone. I need it to fly this high. It needs to have battery last this long, and it's used in a specific agricultural application. And then what model can then do is really look at the components selections that are specific to you. Maybe you can pull out the most relevant data sheets that are used only in such applications while ignoring the noise.

Raymond Yin:
So, to be able to get that much information into a knowledge base or into an AI model, to be able to almost do component down to the component selection level based on a system level parameter.

Nemanja Jokanovic:
Exactly. And that happens with understanding the context of an engineer really and deeply to its core. Because if you understand what they're trying to do, then what the model can do is take all that knowledge and input into consideration, while analyzing data sheets, while analyzing the parameter of each component. It can present the engineer with the most applicable solution at the component level for that particular moment in time. Now, if the model replication evolves, well, maybe now that drone needs to become heavier or you're adding specific camera. When I say, “Hey, I want to add this type of a camera in” because now I need thermal screening on top of everything, that can change. But constantly capturing the context and then evolving the components that go into it for the maximum output needs to be stored and saved within the application, not just one's brain.

Raymond Yin:
And, and not only doing it, literally doing it real time, as you change the parameter, the system parameters, like you said, hey, it needs to be able to do a thermal camera instead of a video camera. That'll completely change your component selection. And being able to do it real time and then storing it real time for engineers to be able to reference immediately.

Nemanja Jokanovic:
Exactly. And going back to our topic of really having the cloud enabled application on top of that - is you can design a beautiful product, be extremely happy with all of that, but then ignore the supply chain.

Raymond Yin:
Oh, that pesky supply chain. So, we actually got to buy the products.

Nemanja Jokanovic:
Well, and I actually know the company that sells such. Yeah, it's called Mouser.

Raymond Yin:
Yeah, we have lots of components.

Nemanja Jokanovic:
Exactly. So, one can do a fantastic job using SnapMagic copilot or continuously doing all these amazing things in their CAD tool. But when it comes to procuring the parts, that's a major aspect also of a lot more need for engineers to be involved in.

Raymond Yin:
Right? Yeah. Because most of the time, an engineer, as long as they choose components on the AVL, they're relatively new, not obsolete. The engineers are generally good with that. And when it gets over to the buyer or whomever, that's a whole other scenario that happens with the project development.

Nemanja Jokanovic:
I mean, I think we evolved enough, and people are in general trained not to design an end-of-life part to their design.

Raymond Yin:
Well, we would hope, but it may have happened a time or two.

Nemanja Jokanovic:
But it does happen. And there are specific companies who serve such market in the supply chain side. So, I do think it's important to really take the supply chain into consideration because, jokes aside, I don't think we don't want to design parts now and come to the end of our design and everything's approved just to realize that that component is not available or has been out of production for a year and a half. And now, I need to go back to my drawing board and find the appropriate alternate. And there's just, again, the whole concept of everything we're talking about is about saving time and doing things more efficiently. So being able to buy components that are continuously available is one of the key aspects of electronic development too, often ignored.

Raymond Yin:
And so, what you're saying is, because we're now cloud enabled, we've got artificial intelligence telling us what components we should be using, that hopefully, all that should be - I mean the supply chain side - any issues should be mitigated as a result of the enablement of the part selection. It sounds like that would be an ideal situation.
Nemanja Jokanovic:
Yeah, continuous from ideation to mass production thread. And they're companies are trying to solve that. I believe we are one of those companies. You have to rely on partners.

Raymond Yin:
So, you guys, I didn't realize this, you guys, SnapMagic is actually, not only are you on the design side, you're also working to enable the supply chain side as well with artificial intelligence and these tools.

Nemanja Jokanovic:
At the time, SnapMagic, when it was originally formed as SnapEDA, we really tried to solve specific need, which was meeting the engineer at the time of a part selection design and enabling them to download their CAD models, the highest quality model in the market they can find at the time of them deciding what part they want to use in their application and probably conquered the market in that aspect.

Raymond Yin:
Yeah. SnapEDA. I mean, Mouser uses SnapEDA. We've had lots of discussions with Natasha and so on. Yeah, we've been working with you guys for years.

Nemanja Jokanovic:
Absolutely. And just watching the evolution of that partnership has been extremely rewarding to me. And there's definitely no intention of pausing anytime soon. We've been having fantastic conversations, but then people often ask, well then what happens? SnapEDA? SnapMagic? And really there's a reason behind the name itself. It's really creating incredible magic for people - engineers, and our partners, and connecting the entire flow beyond part selection all the way to manufacturing. Hence the partnership with Mouser is enabling the pricing and availability of componentry through and applying them into the AI applications of design itself, while enabling the engineer to select the most appropriate component, at the time of that part selection and the design – is solving that.

Raymond Yin:
And so, whether the engineer may or may not realize it, the tool SnapMagic tool is not only picking the technically correct part, but also the best from a supply chain and procurement standpoint as well.

Nemanja Jokanovic:
Because it can take that into consideration.

Raymond Yin:
It's all part of that model.

Nemanja Jokanovic:
Right, yeah. You might have a part that's being suggested to you, but it will say, you know what? It's really actually not available versus, well, yeah, while you research and all that, there's 10,000 of them sitting on the shelf ready to go. It's recommended for design.

Raymond Yin:
The product that was selected is best for technical reasons as well as for procurement reasons.

Nemanja Jokanovic:
Exactly. It needs to be considered because there's no product, without actually having the physical product.

Raymond Yin:
So Nemanja, it sounds like the artificial intelligence and what you guys are doing is pretty much from literally from the conception phase all the way through procurement and actually buying the products and building the products. What other areas do you see artificial intelligence really making an impact in the tool chain?

Nemanja Jokanovic:
Yeah, great question. I think we're scratching the surface, and I think we're actually evolving at a pretty solid pace. But I really think where we are headed when it comes to the AI in particular, is not just using it as a part of reactive suggestion, but more as a coach. Think of it as a senior coach that you can constantly have next to you correcting you or advising you or proposing solutions that you might ignore, contextually. So, I think we're really headed more towards suggestive, proactively suggestive application rather than just a reactive output of at the moment of you requesting something,

Raymond Yin:
So, taking the knowledge within the model, wherever, be it from reference designs or application notes or whatever, and just kind of distilling that whole thing down into a system that can actively and proactively, like you said, advise an engineer on not necessarily what to do, but what to look out for some hints and tips for their designs.

Nemanja Jokanovic:
Yeah, exactly. And I think, at least for me, where this leads me is the question people always ask, well, is AI going to take my job?

Raymond Yin:
I've heard that. Yeah. I mean, I've heard that so many times on different areas of AI. What's your take on it?

Nemanja Jokanovic:
It cannot, of course it cannot. I mean, at the end of the day, it starts in here, and everything we use in today is in service of enabling more of thinking and better thinking. And this is why I think, while AI is going to enable applications to become more efficient, communication, task assignments, supply chain suggestive models, I think really it's going to continue to serve an engineer as a buddy, as a proactive buddy, and continuing to bring potential suggestions that are applicable to a certain design in the context of making the best design solution possible. But I think that's where it's going to be: in the service of engineers throughout that design process.

Raymond Yin:
So, staying a tool, staying an advisor rather than a turnkey press a button, kind of like Star Wars where robots designed robots, none of that. It's always going to stay with the engineer as the center of that design process.

Nemanja Jokanovic:
Yeah, I just don't think we're at a stage yet where we're going to have electronics, design electronics. You're going to have to have human output. 

Raymond Yin:
Interested in more of our conversation with Nemanja Jokanovic? Subscribe to Mouser's Empowering Innovation Together technology series. You'll unlock the concluding episode and exclusive access to technical articles, videos, and more. Sign-up today at mouser.com/empowering-innovation.