Removing Bottlenecks That Burn Budgets — with CognaSIM and CDS

Episode Summary

The episode titled "Removing Bottlenecks That Burn Budgets — with CognaSIM and CDS" delves into the challenges faced by engineers and manufacturers in integrating design and simulation processes. Hosted by Michael Finocchiaro, the conversation features John Zinn from Cognizim and Rhushik Matroja from Cognitive Design Systems (CDS). Cognizim specializes in streamlining the process of moving designs through simulations, particularly focusing on large assemblies and complex systems. CDS develops software that optimizes geometries for manufacturability, cost, carbon footprint, and performance, aiming to reduce iterations between design, simulation, and manufacturing divisions.

During the discussion, key insights were shared regarding the integration of artificial intelligence (AI) in engineering workflows. John Zinn highlighted how AI can simplify the process of assembling complex components into rich assemblies and managing forces, connectivity, and system-level simulations. Rhushik Matroja emphasized the importance of separating data-driven decision-making from traditional IT practices to better manage product lifecycle data. Both guests underscored the potential for AI to significantly enhance engineering productivity by automating tedious tasks and enabling more efficient collaboration across departments.

For PLM and engineering professionals, the key takeaway is that leveraging AI in simulation tools can dramatically reduce design-to-manufacturing bottlenecks, leading to faster time-to-market and cost savings. The integration of advanced software solutions from Cognizim and CDS can help organizations achieve a more seamless transition between design and manufacturing stages, ultimately driving innovation and competitiveness in the industry.

---

Full Transcript

Michael Finocchiaro

So we're live. ⁓ Welcome to the AI Across the Product Lifecycle podcast. This is your host, Michael Finocchiaro. I'm glad to be here and I'm glad to welcome my guest today, John Zinn of CognaSIM and Rhushik Matroja of CDS or Cognitive Design by CDS, if I'm more precise. How are you guys doing today? So ⁓ Rhushik, tell us a little about you and about CDS, the mission you guys are on.

John Zinn, CognaSIM

Fantastic. Thanks for having us.

Rhushik MATROJA

Wonderful. Thank you for having us. Sure, yeah. So, hello everybody, I'm Roushik. I'm CEO and co-founder of Cognitive Design Systems. ⁓ We develop a software called Cognitive Design. It's an engineering software. Our main focus is to optimize the geometries for manufacturability, cost, carbon footprint, and performance. We take a 360 approach on optimization of such geometries to make sure that there are ⁓ less iterations between ⁓ all different various divisions of manufacturing, design, simulation and others. Thank you for having me.

Michael Finocchiaro

absolutely. I think it's great because you're really driving that wedge between the design and simulation side. Whereas John, I think you'll explain to us how you're trying to make it easier to interact with the simulation software.

John Zinn, CognaSIM

Yeah, at Cognizim, I'm the CEO and co-founder of Cognizim. And we really have honed in on a lot of the pain points for engineers in taking their designs and moving it through simulation. And while the industry's made huge advancements in terms of accessibility and tools that can work on the component level, what we see is that ⁓ large organizations really struggle when it comes to assembling these components into ⁓ rich assemblies and looking at systems and analyzing systems and managing ⁓ forces, contact, connectivity, and studying these larger systems. And so we're honed in on using AI to make simulation more accessible and are building a universal simulation interface that's AI driven.

Michael Finocchiaro

Hmm.

John Zinn, CognaSIM

that can connect to multiple simulation platforms and multiple physics and really allow engineers to focus on understanding their products and understanding how to improve product performance instead of wrestling with UIs and figuring out how to connect different tools and how do I do a coupled fluid structure interaction problem and get everything set up and get it to work. And so we can use... ⁓ the power of AI to really streamline that process and just make engineers lives easier, but also make help teams to move faster and do a better job of engineering.

Michael Finocchiaro

That's really awesome. I'm glad to have both of you guys on here and both of them sounded really exciting. What I usually like to start with is talk about this moment we had three, four years ago now when OpenAI came out with GPT-3, which most of its mics have never even heard of GPT-2, but suddenly there was this GPT-3 and now it's pretty legendary. When that happened, when that moment happened, were you... skeptical or you like, ⁓ it sounds like hype or were you like, wow, this is the change. This is the pivot point for engineering. Where did you guys sit on that, John?

John Zinn, CognaSIM

Yeah, it was really interesting. It was fun to play around with at the beginning and it was kind of eye-opening and a little bit inspiring. And to be honest, I wrestled initially with what can you do with this and where can you go with this? And then it took about a year of playing around with it to say, okay, it looks like it's getting mature. It looks like it's developing. It looks like something could be done with this. And the question at that time was, okay, how do you do something with this? And what's exciting is that, you know, the following years we kind of moved into 2024 and especially, you know, all the revolutions around MCP in 2025 just was really the tools and the ability to build in these systems improved. So that was definitely exciting. And now I think,

Michael Finocchiaro

Good.

John Zinn, CognaSIM

The takeaway that I have just looking forward is figuring out ⁓ how do you get it to be performant and accurate and tight to an engineer's standard of quality. And ⁓ the hard lesson for us to learn is that ⁓ prompt engineering and ⁓ MCP knowledge documents aren't enough to guarantee consistent behavior. And so... We're having to go deeper and work on other approaches to say, how can we guarantee consistent behavior that engineers expect and require?

Michael Finocchiaro

Right, so it's not probabilistic but more deterministic, right? ⁓

John Zinn, CognaSIM

Exactly.

Rhushik MATROJA

Yeah, I I echo John on the same thing. And I was impressed by when it came out with the large language models, but I was skeptical on where this could head to and how this could be used in engineering. I believe that this could be a great tool at the beginning that is to create engineering knowledge warehouse or some kind of a database where we can actually gather the know-how and share it. ⁓ I believe that this could also kind of democratize our ⁓ relatively niche field that we are working in the audiences. So they can maybe come into interact with us in a massive way. ⁓ That's what I was my first thoughts about it. The first trials that we did at that moment, it was of course non-conclusive. mean, it was still too early to make something out of it.

Michael Finocchiaro

Because...

Rhushik MATROJA

But like everybody, I I kept the eyes and ears open and on the progress that it made along the months because it's not even years. mean, it looks so far away, but it's actually a months ago. So I never thought it's a trend because the technological foundation that it's based on, it's quite solid. And I think it's a long lasting technological breakthrough. that we are saying, so it's not just a bubble or a hype that's to go away a certain way. It will maybe change how it's been used in the future, like any other technology at the very beginning. And to make sure that we are aligned on that, at CDS, we also started two initiatives regarding the AI. So the first is the AI for productivity, and the second one is the AI for product. because if you understand it well in your daily work, then you can mine also, improve it also and bring it also to the product ⁓ even better. And I can say that the first results are really cool. So very excited.

Michael Finocchiaro

It's pretty amazing. I think CDS started in 2018. So actually before the end of the world and then the restart of the world. And how, how did you start or at what point did you start adding AI into the development cycle? When did you start having enough confidence that the developers could use it and you could get real code out of it?

Rhushik MATROJA

I mean, yeah, so at the CDS, we talk about AI from the very beginning, but it was a different AI that we were talking about. We were talking about more about the machine learning algorithms and some other kind of other AI technologies like CNN and GCN and others. So we were always ⁓ into the field of AI, this whole... But these algorithms were working in the backend of the software and not something that is... interacting with the user on a daily basis. So we were training our models, we were implementing those models into our product and give access to the users. On top of it, from the beginning, we decided that we want to create something which is completely on-premises. So because of the nature of the industry with which we work with. having something where I can bring the data and enrich my where it was difficult. So the whole learning was done on our side. And then we bring the technology ⁓ to the software. yeah, always fan of AI, but now even bigger because of what's happening and what we are looking at.

Michael Finocchiaro

And is it the developers at CDS, are they using cloud code or cursor or stuff like that?

Rhushik MATROJA

yeah, yeah, definitely. So that's a part of the first initiative of the EI for productivity. ⁓ it has been game changer in the speed of execution that we can achieve. Thanks to this, of course, at the very beginning, we were very skeptical about the quality of the code and would that be enough to bring it and there will not be any bugs after and everything. But we see that more and more we are progressing. The quality of code is becoming more reliable. we can define, but the engineers, the developers, their role have transformed in a way, because instead of typing the code manually and debugging it manually, now they have an assistant. Everybody has a secretary, kind of, so everybody's happy in a way, and they are able to achieve and do some more productive work, which is more on the side of like...

Michael Finocchiaro

to see that you need some more, let's do one which is more on the side of

Rhushik MATROJA

the architecture, what thought process which is going behind the algorithms, rather than just

Michael Finocchiaro

the architecture, what's the most basic design algorithm rather than...

Rhushik MATROJA

typing the code and fixing where the bug is, where I forgot the comma and where I forgot the, where there is a spelling mistake, for example.

Michael Finocchiaro

you John, do you have a similar experience? I you started in 2024, correct?

John Zinn, CognaSIM

Yeah, so we definitely were able to build upon ⁓ all the power of Transformers that was coming to market and starting to mature. And it was exciting to see ⁓ how Cursor and how Cloud Code were being developed and their capability. And it kind of was an inspiration and a model for us to say, look, this can happen. And looking at how large organizations were skeptical of AI and how they began to embrace it. And just like Rhushik said, there's really a change in workflow. You get an assistant and it can help you write code and it does a pretty good job. But the software engineer's role now changes to be, you you need to be a better code reviewer and you're reviewing code that AI is helping you to generate. And you need to think. a little bit longer and harder about the architecture. And these coding assistants will also help you develop the architecture. But one of the interesting things about AI, and I do see it as a transformational technology, and kind of plays into this roadmap of education and jobs, where it goes, is that the thing that people have to be

Michael Finocchiaro

The thing that people have to be

John Zinn, CognaSIM

cognizant of AI and know what it does well and know what it doesn't do so well and know how to communicate clearly to it to get the intended outcome.

Michael Finocchiaro

cognizant of AI and know what it does well and know what it doesn't do so well and know how to communicate clearly to it to get it intended.

John Zinn, CognaSIM

Words matter and ⁓ something it's kind of like ⁓ talking to your children. You say something you think that what I said was very clear. They heard what you said and interpreted it slightly differently and

Michael Finocchiaro

It's kind of like talking to your children. You say something, you think that what I said was very clear. They heard what you said and interpreted it slightly differently.

John Zinn, CognaSIM

So it's forced me to kind of look at am I being clear? Am I being precise in my instructions and the other thing that's interesting is that

Michael Finocchiaro

And so of course, you need to kind of look at, am I being clear? Am I being precise in my instructions? And the other thing that's interesting is that...

John Zinn, CognaSIM

There's a lot of power there but if you don't have the creativity to ask the right questions or to challenge it in the right way to think big enough or to do more for you then it can look like a weak assistant for you instead of being a partner working with you. And so the workforce needs to learn how to work with how to do it responsibly. Understand that you're still a human and you're still in charge and you're still responsible at the end of the day and you need to take ownership to but also to say, you know, challenge the AI and think bigger and give it

Michael Finocchiaro

the to meet all of our needs.

John Zinn, CognaSIM

larger, more complex problems to solve and see how it can do that for you. And more than likely, you'll be amazed, you know, with the companies that are really leveraging the technology and implementing it well.

Michael Finocchiaro

Very cool. I was wondering too, so like in terms of the development process, are you guys, how are you guys building the knowledge base on the engineering side? Are you guys using RAG? Are you using, are you thinking about developing your own LMM or LMM as Mauer of Leo AI talks about? I mean, how are you basically capitalizing on the engineering and simulation knowledge to help the engineer figure out the, you know, that that fill it's not going to work. And actually we ought to do it this way. Rhushik, do you have some input there from a CDS point of view?

Rhushik MATROJA

Yeah, yeah, definitely. mean, on the, you mentioned, the engineering drag, ⁓ that's something that has to be part of, believe in all engineering software companies that would be very, very important because ⁓ first of all, having ⁓ the context, because now that's the biggest word, like understanding the context of the problem and keeping the context alive. ⁓ on same time updating the knowledge base so it can also maybe find the right patterns ⁓ and then implement it over the years or over the months of using it that, these patterns are coming up more often than, we should propose or the software should propose it automatically. That would be there. But that's the thing. mean, this is the fundamental change on how the software works. It's no more ⁓ one way of like, OK, the orders are given and you have some specific sets of orders give you a specific sets of solutions. ⁓ Instead of that, this could be something that could be much more ⁓ improved, evolutionary. ⁓ And that's where we are heading to, definitely.

Michael Finocchiaro

Awesome. Just so the folks in the audience feel free to ask questions. I'm looking at the chat when I look over here. So feel free to ask questions to Rhushik and John live. John, did you want to react about that ⁓ last question?

John Zinn, CognaSIM

Yeah, our process is very similar. And in many ways, we kind of look at AI as being a student. And we're teaching it about engineering and engineering principles. And then our focus is really heavily pivoted towards testing and refinement to say, here's the problem. here's the answer, go solve this and did you get the right answer and understand how do we score it? It's a C, it's a B +, it's a A. And teaching it to say, now we're scoring you based on this and we're looking back at ourselves as the teacher and saying, were we not clear in what we were teaching you to do or are you not being a good student and you're not listening? to what we're saying. And so it's this process of working with it. And Rhushik also brought up context. Context is king. And how you manage context is really important in ⁓ managing the response that the AI gets, because it's heavily biased towards what you've been talking about the most recently. So if you... do a hard change of thought and you like bounce back and say, ⁓ this other thing, and you're not explaining your thought process of how did you get from point A to point B and like, what are we doing now? ⁓ That can create situations where it gets confused and may misinterpret your intentions. ⁓ It kind of thinks you're still working on the thing that you were last working on. so if you're not clear and it's pretty easy to... to lay out exactly what you want to do. ⁓ And I think the wonderful thing is that ⁓ we're fortunate to be working with engineers who kind of think in an orderly, structured thought process and are very good at working with these systems and feeding information to them.

Michael Finocchiaro

Yeah, I found that a lot of times the AIs will just kind of go down a blind alley and then they get stuck there. can't have them turn around and start over again. ⁓

Rhushik MATROJA

Yeah, I that's a great point. And I echo John on that one thing that he mentioned about the determinism. That is also very important in the whole process that if your software is based on something more concrete, more deterministic approach, you can have AI layers to orchestrate various different algorithms. Having the certainty of answers, having the context alive, that would be the key to real engineering because that's differentiated engineering and art that, okay, if it's not repeatable, then it can become quickly a piece of art that, okay, yeah, okay, this is a very unique piece, but I mean, we need something which is repeatable. And that would be the biggest challenge in the coming months.

Michael Finocchiaro

I mean, we've talked a lot about, ⁓ not us in this conversation yet, but a lot about agentic AI and the... But what I found the problem is that it's always trying to make you smile and make you happy. And I feel like there needs to be a MyMinutes agent who's sort of spewing doubt. I don't know, I'm not really sure about this. ⁓ It just makes you think also, I think that might be a whole field of AI ops we're gonna need. Because we need to manage the verifiable facts in a conversation, separate those out from the frustration on the keyboard saying, God damn it, didn't you hear what I said? I'm sure we all do that to our GPTs when we're trying to get frustrated with them. I think it's complicated because you've got to separate the stuff that's actually real and works from just the conversational stuff. And that's not always easy to get to,

Rhushik MATROJA

Yeah.

John Zinn, CognaSIM

Yeah, interesting.

Rhushik MATROJA

I heard a new word about it. That's a swarm bots that are like a swarm agents, agentic swarms, or I don't know how they call it, but yeah, that allows you to like have various different roles. And I think that's like having this critical analysis AI that does the critical analysis or have that kind of mindset ⁓ would be important.

John Zinn, CognaSIM

Yeah, there's a...

Michael Finocchiaro

swarm but

Rhushik MATROJA

We are already talking about AI as somebody who is there, but yeah, that's how we are getting familiar with that.

Michael Finocchiaro

John, had something to say, I think.

John Zinn, CognaSIM

Yeah, we've played around with that and have really considered a concept of having an AI as an independent verification AI. And do you need an AI to watch the AI and to hold the AI accountable? And the interesting thing, it's challenging, it's interesting and a challenging problem in the sense of if you need to feed the same context to the AI and if there's a probabilistic element.

Michael Finocchiaro

And if there's a probability element

John Zinn, CognaSIM

to it, then there is a chance that one says go slightly left and one says go slightly right and they're both going in the same direction but they're taking

Michael Finocchiaro

to it, then there is a chance that one says go slightly left and one says go slightly right, both going in the same direction, but they're...

John Zinn, CognaSIM

a slightly different path ⁓ and there could be some disagreement over that. And then the other thing is if you give it too much context to do the validation to say here's everything, making sure that it doesn't fall into the same hole or fall into the same trap. How do you get this independent set of eyes? And so it's almost, what we've concluded is there's a need to present limited context to look at the last statement by itself and evaluate it by itself and evaluate the efficiency, effectiveness, and accuracy of the AI, the primary AI, the agentic AI that's doing the operation.

Michael Finocchiaro

What we've concluded is that we need to present a context to look at the last statement by itself, evaluate it by itself, and evaluate the efficiency, and accuracy of the primary AI, the identity AI that's due to the operation. I was also curious too. suppose you guys use some version or adaptation of waterfall or agile in your environments. And AI seems to be changing that as well, right? I mean, do sprints really matter anymore? People with your agent, you're sort of doing things on your own anyway. How is the role of a developer and then the role in agile or waterfall? mean, are you guys, well, first of all, are you guys using one or the other or a blend? And then how is AI sort of changing, morphing that? into something new.

John Zinn, CognaSIM

We're using pretty much an agile process and have basically morphed it to meet our needs and to move at the pace that we're developing so that agile doesn't become a rigid framework that constrains developers into this. And so the challenge that we face is looking at problems and tackling the discrete features that need to be built. while at the same time understanding architectural features and foundational ⁓ capabilities that need to be designed and architected and tested and laid out and rolled into the product and how do we do that and do that quickly. So we've kind of morphed our own agile process to make it flexible for how we develop.

Michael Finocchiaro

And at Rusek, maybe you guys already have an AI scrum leader at CDS.

Rhushik MATROJA

⁓ Yeah, I mean, that's what I was telling you that the role of developers and engineers is changing quite a lot. So I think currently we are also like, like John mentioned, we are also on the Agile method with the traditional, I would say, approach that we are trying to modify it or morph it to demand also the AI agents. But... ⁓ What I'm seeing is that, it has to change also on how people interact with AI and how everybody using various different AIs, how their AIs also can communicate and create something which which makes, how can I say, which is more predictable and usable and that matches each other's programs.

Michael Finocchiaro

seeing the aesthetic as it came also. How people interact with AI and how everybody using their systems, AI, how they are able to communicate and create something different.

Rhushik MATROJA

That is the biggest challenge. And for that, we will need a very soon an AI scrum leader. Definitely.

Michael Finocchiaro

Do you find that developers work better together because of AI or just differently perhaps? I mean, because now it seems like

Rhushik MATROJA

Alright.

Michael Finocchiaro

The guy you talked to is not the developer sitting next to you, it's the co-pilot. So is there sort of an isolation from a human point of view because of the use of AI or you still see developers going to talk to their friends and chatting and stuff?

Rhushik MATROJA

It's a cool platform. I think the discussions have changed. I would say that they are not talking about, okay, how to resolve this bug or that bug with which ⁓ architecture to use, et cetera, et It's more on a higher level, ⁓ which is great thing because they are also evolving in their career also in this way, that they are also talking about they are thinking a little bit larger vision of the product. ⁓ which is a great thing. But I think, yeah, this increased the collaboration rather than reducing it, definitely.

Michael Finocchiaro

That's great. You have the same experience, John, you're nodding.

John Zinn, CognaSIM

Yeah, yeah, very much very similar. And I think ⁓ part of it is that the touch points, the cadence between touch points is faster. So as we develop faster, and you're using AI to build a function, now you're talking about the next feature and what it should do and how it should work and how it should be implemented. And so you're not going off and coding for three days and refining and then going through code reviews and ⁓ doing revisions, the cadence of the coding is picking up. We're able to develop features faster, and that does lead to more touch points between engineers because they're not just sectioned off working on their own thing in isolation.

Michael Finocchiaro

Does that mean that us as the developers are trying to become more multidisciplinary? Like maybe some are very only doing UI and other ones are only doing control and now everybody's sort of expected to do a little of everything, right?

John Zinn, CognaSIM

Absolutely. Yeah. Like all of our engineers know how to work and how to build UI features and they don't have to have the best JavaScript skills. For example, it's like, you know, go create this thing for me and it builds it.

Rhushik MATROJA

Absolutely.

Michael Finocchiaro

So,

Rhushik MATROJA

I think there is also more enthusiasm also on the staff in general, would say, not just developers and coders. There are so many new things coming over, all the applications based on those LLMs that are coming over. So ⁓ educating each other on the recent trends, recent new things that comes over. That also creates a lot of harmony between them. And ⁓ it broke down a little bit of the barriers that that might have in the past that, OK, I'm working in my division and not in yours. Because now you can actually share a lot of things between them and teach each other that, I learned this and that. ⁓ So it's pretty good in this way.

Michael Finocchiaro

I'm working in my division and my EOS because now we share a lot of things with them and so they're better than numbers and in fact they're pretty good. Well, before we move on to my next question, there is a question in the chat, is pretty interesting. So Muaz Al Akadzi says, I want to get more technical here. Do you have insights on the AI usage of foundation? models versus the widespread LLMs like ⁓ Anthropic Cloud and ⁓ OpenAI's ChatGBT. We sort of touched on that earlier, but do you see customers asking for their own fat international models so that they don't get the code eventually, their IP shows up in somebody else's chat on ChatGBT?

John Zinn, CognaSIM

Yeah, I think there's a lot of concern around that. And it's a challenging problem, I think, in the industry right now, which is to say, can the specialists develop faster than some of the larger LLMs that are bigger and have more resources? And so it's a little bit risky right now in the industry to pivot and to go off and to build your own foundational model because it can very quickly become outdated and general advancements are pushing the frontier so fast ⁓ that you can get left with old technology or stale technology or have limited capabilities as a result. And so ⁓

Michael Finocchiaro

Betamax, you can Betamax yourself.

John Zinn, CognaSIM

Yeah, exactly. So the challenge is really understanding customers' concerns, working with them, and providing solutions, while also giving them the performance that they require and staying at the cutting edge. Because everything's really changing very quickly. And at the end of the day, it's best for the customer, like our customers, it's best for them to have the most recent advancements. on.

Rhushik MATROJA

Yeah, I found the same way. mean, creating our own LLM, mean, that would be very expensive, but on same time, we won't be as competitive as people who are already specialized in that. having said that, I believe that, I mean, yeah, of course, plugging to other LLMs or the larger models would be one way to see. But I also saw that some of our larger customers, OEMs, they are developing their own maybe modified version of LLM or they have their internal system. And I see that, okay, that is more interesting and that's where we have to be able to stay agnostic towards all this different various LLMs so we can plug to a specific LLM based on the customer requirement would be necessary.

Michael Finocchiaro

But I also saw that some of our large companies in the area, they are developing their own, maybe modified version of that. I think that is more interesting, that's why we have the ability to stay agnostic towards all this different elements, and can help to close that particular element based on the customer requirements.

Rhushik MATROJA

Because as a specialist software, we are not going to bring as much value to the LLM itself, but we're going to bring value on how to use our algorithms better using those LLMs.

John Zinn, CognaSIM

And I'll just add one thing. ⁓ It's interesting, Rhushik, that you mentioned that you started with machine learning. And in our approach, and one of the things that we're thinking about is understanding what do LLMs do well, and then saying, what does machine learning do well, and blending those approaches. And so there's power that comes from

Michael Finocchiaro

So speaking of, oh guys, so John, don't want to interrupt you.

John Zinn, CognaSIM

⁓ not taking the easy way out and not just leaning on the LLM to do everything by itself, but there's power that comes with taking a more holistic view towards AI and integrating machine learning strategically where it can give you the most power.

Michael Finocchiaro

Awesome, thank you for that. Let's move on to the next section, which is about when someone is using Cognisim and CDS, where are the touch points at the AI? Are they actually seeing an agent that's doing stuff that's AI-ish? Is there a little chat bot on the side, or is it more the plumbing, the ML-informed optimizations underneath the covers? Where are the touch points between the user of the software and the AI that you're deploying?

John Zinn, CognaSIM

Yeah, we've built an on-prem solution that is a standalone utility that then connects with and interacts directly with various CAE and FEA software. And so we find that to be really effective in the sense that our interface is, it looks a lot like a chatbot, but it gives us the power to ⁓ add features and add capabilities that ⁓ that aren't readily available through a chatbot. And so we're really building an ecosystem and an environment that the user interacts with. And we can keep it very simple and streamlined, but as they work with it, one of our key ⁓ guiding principles is that we don't want to build a black box. So we don't want it to go off and execute something and then deliver you this result on the silver platter and say, Isn't it amazing? And do you like it? So we're very transparent in our workflows. So as you interact with our AI agent, it's going and it's performing the functions that the user requested in Ansys. And then you see it actually building a simulation in Ansys and completing the tasks. And it's powerful in that way because they get the visibility. And somebody who never had used Ansys before can see. it created this. ⁓ it did this. It added this here. And they can start to get a feel for the simulation software and start to get comfortable with it. So we're not taking the expert engineers out of that platform. We're just augmenting and accelerating their workflows.

Michael Finocchiaro

But there's a danger if they figure out how do it right, they won't need cognizant though, right? ⁓ Don't want to show too much, okay.

John Zinn, CognaSIM

Yeah, we're not worried about that. Yeah, we make it faster and easier to do just about anything they want to do. ⁓

Michael Finocchiaro

How about you, Ruchika, as I work for Cognitive Design?

Rhushik MATROJA

So we are ⁓ the chatbot and for applications, it could be one way of doing it. On our side, as we are a design software and we need interaction, quite a lot of interaction with the 3D models, but also with the whole workflow that is created to optimize that 3D model for performance, for manufacturability, for cost, for carbon footprint. So all these modifications requires a lot of interaction and having only chatbot would not be enough for in our use case specifically. So we believe that bringing our whole model that we are constructing, can't say very much about it, but it would be the different way of interaction with the 3D models. It may not be the chatbot. It would be more natural in a way to interact with the 3D models that would be there. We'll be using LLMs and other things in various ways. But what we want to focus on is to bring a fundamental change on how the engineer designs a product, ⁓ how it communicates with the software in the best way possible. And there is a real interaction between the software and the engineers so that it can evolve in the right direction. So that's what we are building at the moment. So yeah, that's why I can't a lot about it.

Michael Finocchiaro

That's okay. But I recall when I spoke to you or maybe with Vincent that you guys developed your own graphics kernel, that you're not using Parasolid or ⁓ OpenCascade or Asus, which is pretty interesting too, right?

Rhushik MATROJA

Sorry, your voice got cut a little. Yeah, we don't use it. Yeah.

Michael Finocchiaro

I'm sorry. You built your own graphics kernel, which is a huge undertaking.

Rhushik MATROJA

Yes, exactly. We are based on our implicit modeling kernel. So we have complete control over our geometry engine. ⁓ And the interaction that happens with the could be very different from select.

Michael Finocchiaro

He had some networking problems. You back?

Rhushik MATROJA

Yeah. Sorry. I'm back. Can you hear me?

Michael Finocchiaro

We can't see you but we can hear you.

Rhushik MATROJA

Okay, sorry about that. Let me start again.

Michael Finocchiaro

It's okay.

Rhushik MATROJA

Sorry about that. Yeah. It's an internet connection problem. But yeah, on basis of implicit modeling kernel that we have, we added the functionalities of basic FEA, so called static analysis, mechanical and thermal. We have also manufacturing analysis, cost analysis, and carbon footprint analysis. So all this together requires a lot of flexibility. ⁓ in the interaction side and that's what we are trying to create.

Michael Finocchiaro

Okay, just in case you didn't know Rishikesh, we don't have video of you. We do hear you though. Okay, no problem. And now that we didn't want to see you, but that's pretty cool. So now we're in 2026. We just turned the calendar over and at the end of 2026, we had yet another revolution with this crazy mold bot, clod bot, yet another very...

Rhushik MATROJA

I will

John Zinn, CognaSIM

Thank

Michael Finocchiaro

autonomous, loosey goosey kind of use WhatsApp and talk to your AI and just create stuff. I talked to one founder last week that said he was talking to his phone and his 3D printer just starts making stuff. just like a replicator, right? We're back already. We're already there. Where do you, I mean, how has your opinion of AI changed, John? You came in, you said you were already pretty enthusiastic, but at the same time you wanted to see a little bit of wait and see. Well, you've had the four years, you're already using it to a game, but are you still as bullish or you still have some reservations? And where do you think it's going, Frank?

John Zinn, CognaSIM

Yeah, I think that ⁓ I'm really excited about the power that it brings when AI is clicking and it's doing everything that the users want. And the other thing that I'm bullish on is the power of AI for scalability. And so the way that we started building systems wasn't the most scalable. And so you work on technology, you develop an approach. And I think the other thing is that AI is making the development of AI based systems faster and more efficient. And so that's, I think, exciting. And the challenge is really, I think it will be a challenge separating out the signal from the noise of a lot of conversations about things that can be done with AI and demonstrating a real user focused and productivity ⁓ emphasizing capability of AI, not just that it can do something cool, but here's how it matters and here's the tangible outcomes and measurable KPIs on how it's making a positive impact in your organization. And so some companies, if we step back and we look at the broader AI in the engineering space and AI in the simulation space, there are a lot of companies that are focused on surrogate models and other approaches such as that. that take a really large investment and don't yield immediate ROI. And so our focus is to say that companies need to see ROI, they need to see it faster. And so we're excited just to use productivity enhancement to do that. And so I'm very bullish on AI and where it's going and ⁓ the advancements.

Michael Finocchiaro

and where it's going and the advances.

John Zinn, CognaSIM

But you have to be flexible and adaptive because you can't get locked into one system or one approach. And so you have to be really staying at the cutting edge and understanding what's happening in the market and what's the new approach and how do I pivot and how

Michael Finocchiaro

But you have to be flexible and adaptive because you can't get lost into one system or one person. So you have to be really staying at the cutting edge and understanding what's happening in the market and what's the solution. How do I say that?

John Zinn, CognaSIM

do I integrate this? And that can be challenged because it's so fluid and it's moving so quickly.

Michael Finocchiaro

and that can be talented. Yeah, so you gotta watch the A Across the Product Lifecycle podcast to see the latest. Rhushik, we still can't see you Rhushik, are you still there?

John Zinn, CognaSIM

Exactly. Don't miss an episode. You'll get left behind.

Rhushik MATROJA

Yeah, I'm still there. I'm still listening to it. It's very interesting. Sorry, buddy. Sorry, I don't have the control over the camera in this app. Yeah, let's go.

Michael Finocchiaro

you That's okay. But do you want to talk about how your perception of AI has changed over last four years?

Rhushik MATROJA

Yeah, definitely. mean, perception of AI. as I mentioned, we were from the beginning, we started the company from the beginning, we were working on the ML models and everything. So for AI for us, it was always part of our journey. I'm a big fan of what's happening at the moment. It's incredible to be alive as a mechanic. But we need to put also customer and their data at the center of our decision making process. Security and accountability over the customer data would be our primary focus while implementing AI in our tech. moving forward, we see that ⁓ it has ⁓ lowered the boundary of lot of technological ⁓ fields that were kind of safeguarded by big players in the past. Now it's becoming more of a level playing field. So people who understands the best, the customer, who can dream the biggest, and who can implement it the fastest, would be the big winners out of this. So turn 36, it's going to be very, very busy year.

Michael Finocchiaro

So, have the defense of that. We can use the business to do the, mean, the business, and then just the process. We find the business out of it. I think that it takes us very, quickly.

John Zinn, CognaSIM

Mm-hmm.

Michael Finocchiaro

Just before we move on to the digital maturity question, do you guys have any advice for the younger engineers that might be watching in terms of things to focus on in order to not have AI steal your job, which seems to be a common source of anxiety for a lot of folks.

John Zinn, CognaSIM

I would say that ⁓ technology changes and the best way is to not be skeptical of technology or not to ⁓ be dismissive of it because if you do that, it'll move on and you'll look outdated. And so the real challenge that I would say to new engineers in the workforce, whether you're in software development or if you're in doing product development. is to embrace AI and understand AI and become an AI master. You don't need to start your career as a green engineer that needs to learn things and doesn't know anything. You can ask questions, very specific questions about how do I do this thing? And you can go and get information that you need and on the job training and ⁓ task specific.

Michael Finocchiaro

Right.

John Zinn, CognaSIM

guidance using AI. And so I would say lean in and do it, but also learn how to evaluate and weigh what it's telling you and say, you know, is this adding up and does this make sense? And go back to your engineering principles and use those as your foundation and evaluate AI in relationship to that. And so you have to ask the right questions. You have to think the right way and not just ask questions and take the answer that you get, but you have to know how to challenge it and know how to go deeper and get a more complete understanding instead of just a surface knowledge.

Michael Finocchiaro

He's back. There you go, swoosh. I was wondering if I had to put a stuffed animal or something in your spot. So what did you want to say about to the younger generation?

John Zinn, CognaSIM

Mm-hmm.

Rhushik MATROJA

back. Sorry.

John Zinn, CognaSIM

Nice.

Rhushik MATROJA

Yeah, mean, definitely to young engineers. I really believe that we are going back in a way in way how engineering was done. Fundamental physics, mathematics, the knowledge about it would be even more important to be able to judge if the answers that are coming over from the AI are relevant, are are correct. And as the whole engineering role is evolving, it's exciting time to be an engineer in this field, because it's not taking away any jobs. On the contrary, it's taking away all the painful tasks that we were doing in the past for all all this 30 years or some that has that will go away. So for me, it's a good thing that AI has arrived.

John Zinn, CognaSIM

Mm-hmm.

Michael Finocchiaro

The Treasury.

Rhushik MATROJA

But on the same time, that doesn't mean that whatever he says, it's whatever the AI can produce is correct. We need to be careful about the judgment and how to do that. Having a bad understanding of mechanics, of physics, of thermodynamics, of mathematics would be key to do so. So yeah. ⁓

Michael Finocchiaro

Vigilant. Thank you. ⁓ So let's just shift gears because we've got another 15 minutes. ⁓ I like to talk about digital maturity in the enterprise because you guys are going, obviously not going into green fields, right? You're going to companies that already have ⁓ engineers that are already developing products. It's pretty rare that someone starting out is going to go immediately to a startup, right? They're going to have some legacy. ⁓ And I think that, so first by the question and a quick response would be, when you are going to your customers that are using CDS and Cognizim, you get a feeling, ⁓ let's say on a scale of one to five, how mature they are. Like one I would think of is, know, almost everything is still email and Excel. Five is like agentic, adaptive, you know, digital twins that nobody's at five, right? Almost nobody. And probably nobody's at four, but in your experience, Rhushik, where are most of your customers? ⁓ Closer to one, closer to three, somewhere in the middle.

Rhushik MATROJA

I think they are more close to one than five years for sure. That's for sure. But since the beginning of the year, I have seen changes. I have seen initiatives in the large organizations that we work with in aerospace industry in Western Europe. ⁓ That they have created initiative to understand what's going on with the whole AI hype. They have special task forces. have... ⁓

Michael Finocchiaro

I'm Yeah.

Rhushik MATROJA

the budget allocated for the specific requirements. And we have more and more requirement that, okay, yeah, I mean, are you using AI? How are you using? The questions are more relevant. They are trying to understand, they are trying to position the company because I think they have understood that if they miss out on this, this opportunity, ⁓ the competition could take leap forward. ⁓ And this would be considerable to catch up in the future. So ⁓ it's amazing to see that even the large organizations, which was very difficult to move three years back to make them understand that, OK, yeah, use generative design algorithms. They were very skeptical about it. But now suddenly, Cloud Code, yeah, OK, why not? We are trying it. OK, that's surprising. But OK, let's go for it. This is interesting.

Michael Finocchiaro

John.

John Zinn, CognaSIM

I see something similar and I would say that a lot of the customers that we talk with fall in like the one to two category. And there's a slight difference in that companies that are doing hardware development that have a software component, a heavy software component to their product ⁓ have taken advantage of this kind of leading edge of software development coding agents. And they now see the power of AI. and they can see it ⁓ making an impact. You're going from like early concept, early adopters to within two years, widespread adoption in the software development field and seeing impacts and ⁓ companies really feeling that if like as Rhushik said, if I'm not using this, I'll get left behind. So that's happened in the software space and companies that have exposure to that, that have a software component to what they're building now are thinking about how can I get the same wins in hardware? Because hardware traditionally, in terms of, ⁓ if we just looked at, we talked earlier about agile processes. And so that's a very old concept in software and is just making its way into some hardware development teams. So hardware can be five to 10 years behind. And when it comes to the adoption around AI,

Michael Finocchiaro

when it comes to the adoption around AI,

John Zinn, CognaSIM

It seems like hardware is starting to catch a vision and is catching up. It is about right now, about two to two and a half years behind where software is. So that's exciting. And I would say that companies that don't have a software component are a little bit closer to the one of doing things the traditional way. This has always worked for us. Risk for them really is not having the opportunity to see the light

Michael Finocchiaro

it's been like hardware is starting to catch a vision and it's catching up. It is about, right now about two two and a half years behind where software is. So that's exciting. And I would say that companies that don't have a public money on it are a little bit closer to the one of doing things in a way that you've always worked your way up. A risk for them really is not having the opportunity to see the light.

John Zinn, CognaSIM

in how AI is impacting the software development field.

Michael Finocchiaro

in how AI is impacting our development skills

John Zinn, CognaSIM

⁓ means that they may stay in the dark and remain skeptical about AI's capabilities for hardware development.

Michael Finocchiaro

means that they may start to feel skeptical about AI capabilities. And have you guys seen any cases where adopting an awesome AI-powered solution like CDS or Cognisim has created some kind of ripple effect or an aha moment or an epiphany for management to say, you know, if we really worked on this digital maturity stuff and we broke silos and we actually work together as one company rather than a bunch of separated and hate, business units that hate each other. my gosh, we could be so much better. Have you actually seen that in the real world? Rusek, you want to start with that one?

John Zinn, CognaSIM

Yeah, I do.

Rhushik MATROJA

Yeah, yeah, definitely. mean, this is the good part of this job that we finally see it, ⁓ that people are realizing it, large organizations are realizing it, that, OK, ⁓ this is where we need to join forces rather than ⁓ working in silos because our competition is not internal, but it's external. And the knowledge platform that we need to begin to share the know-how of an engineer ⁓ is key to oral product development acceleration. So that's the message that we are conveying to them. And it's much more receptive last two years than the years before, definitely.

Michael Finocchiaro

You said the same experience,

John Zinn, CognaSIM

say in very small windows and very small pockets because ⁓ what we're seeing is that it really has to start from the top and it has to be leadership driven and it requires good corporate discipline to say it's not the fact that AI can improve collaboration which is one of our one of the core foundations of our platform is that it really comes into the human element and human nature and says, like, do humans want to tear down these walls? Because the silos that exist were built by humans. ⁓ And they're built to protect groups and to protect organizations and to structure things and to have discipline and to organize and to have formal channels of communication. And so...

Michael Finocchiaro

tear down this wall because the silos that exist were built by humans.

John Zinn, CognaSIM

companies that want to move faster and want a more fluid environment and see the benefits of that, it really has to come from leadership above and that has to be propagated to every level. now the tools are there to really transform how businesses think, how to create a true digital thread throughout product lifecycle. But I'm optimistic that more and more companies will get there, but

Michael Finocchiaro

but I'm optimistic that more more companies will get there,

John Zinn, CognaSIM

It really starts with human nature and people that have a big vision and want to implement it.

Michael Finocchiaro

it really starts to get really rough. How is this different? I don't want to get too sidetracked before we conclude, but I'm wondering too, how many of the companies you guys work with have separated the data from IT? Because I think that's one of the problems too, is if you keep it all under the same hat, it doesn't really work because data is not an IT solution. Data is a whole philosophy, right? And you need data stewards and data custodians and data owners and data scientists in order to really become a digital enterprise. How many, like what percentage of the companies you worked with had a separate line of reporting for data and IT and which, as opposed to having just one hat.

Rhushik MATROJA

On my side, not many, to be honest. I would love that they would do it. This is a very important topic that you mentioned, Michael, and I think it's still not enough understood. They understand that the data-driven ⁓ decision-making would change dramatically, but it hasn't been separated and treated differently from before, from my perspective.

Michael Finocchiaro

Mm. Same with you, John.

Rhushik MATROJA

Maybe I'm wrong.

John Zinn, CognaSIM

I agree and I would say that ⁓ in the better organizations, data gets its own subgroup within IT, but IT is still the gatekeeper. ⁓ And so, yeah, it's a really interesting avenue and important for companies to consider that as managing this data across product lifecycle changes.

Michael Finocchiaro

I get the impression that on the, and I'll talk about this in another podcast, obviously, but I get the impression maybe on the manufacturing side, there's been more of a push for that because of UNS, because of the way we've been able to look at machine data and it's OT, it's not IT, right? So there's already been a bit of a separation. So hopefully that'll come into our side, into the design side of that, engineering side of the house. So before we wrap up, I just wanted to second with you guys in terms of like, I was going mention that I've got a threaded series of conferences coming up. There'll be one in Warwick, UK on the 25th of March. think Rhushik, you said that CDS was going to be there anyway. So if you want to meet Rhushik and the CDS team, they'll be at Warwick. ⁓ On the 13th of April, I'll be down in Miami. think, John, you said you may come down for a cafe colada with me. So you can come to Miami and meet John. Are there other events where people can meet you guys in the next couple of months?

John Zinn, CognaSIM

Yep. Yeah.

Michael Finocchiaro

Where can they meet, Congress and CDS?

Rhushik MATROJA

So yeah, and we will be present at, as you mentioned, the Develop3D Live. We'll be there also at AMUG, A-M-U-G. It's in Reno, Nevada. ⁓ CDFM will be there. We'll be there also for Farnborough, and we'll be there for ⁓ a couple of aerospace events in ⁓ Western Europe.

Michael Finocchiaro

Develop3D. And CD Fam, are you going to go to Barcelona? Awesome.

Rhushik MATROJA

We'll try to be also on the Japan side for manufacturing world ⁓ later this year.

Michael Finocchiaro

I might do a threaded over there. So hopefully we'll get together over in Japan to have a you, John.

Rhushik MATROJA

⁓ sure.

John Zinn, CognaSIM

We'll be at several ANSA simulation world events. There are some local events throughout the country on the West Coast ⁓ and also in Boston. And we'll be at the PTC Developers Conference. And we'll also be at several NAFEMS events this spring, early this year.

Michael Finocchiaro

Nice. Awesome. Well, I really enjoyed this conversation today, guys. Thank you so much. Thanks to the audience. Thanks for the two or three questions we got in the chat. Yeah, and I'm looking forward to meeting you guys both in person this year. That's really awesome. well, anyway, I just wanted to say thank you to everybody. Do you guys want to say goodbye? Parting words.

John Zinn, CognaSIM

Thanks for organizing. Yeah, thank you for your time and...

Rhushik MATROJA

Thank you very much everybody for attending and thank you Michael for hosting.

Michael Finocchiaro

Yeah, well thank you. And I'll be back next week. I think I've got four webcasts for you guys next week. A lot of manufacturing, a little bit of engineering and simulation. ⁓ But we'll catch you on the next episode. Thanks everybody.

John Zinn, CognaSIM

Yes, thank you. Fantastic. Thank you.

Rhushik MATROJA

Thank you.

Michael Finocchiaro

That was awesome.