Null to Infinity: AI-Driven Engineering Workflows — with Nullspace and InfinitForm

Episode Summary

The episode titled "Null to Infinity: AI-Driven Engineering Workflows — with Nullspace and InfinitForm" delves into the integration of artificial intelligence (AI) in engineering workflows through the perspectives of Masha Petrova from Nullspace and Michael Bogomolny from Infinite Form. Nullspace specializes in electromagnetic simulation software, while Infinite Form redefines design for performance and manufacturability by generating designs fully automatically to consider both manufacturability and performance. Both companies are leveraging AI to revolutionize their respective fields, with Bogomolny noting that the current tools fall short of expectations due to disconnected workflows.

During the discussion, key insights emerged regarding the potential of AI in manufacturing. Bogomolny emphasized that while there have been previous attempts at revolutionizing the manufacturing space, a unified solution and platform are now needed to connect various pieces together. Petrova highlighted the importance of CAD cleanup features in their upcoming release, which will enhance design efficiency. Both companies see AI as a catalyst for unifying fragmented processes and automating workflows.

For PLM and engineering professionals, the episode underscores the transformative potential of integrating AI into existing workflows. By embracing these new tools, engineers can achieve more efficient designs that take manufacturing constraints into account, leading to improved product development cycles and enhanced overall performance in the manufacturing sector.

---

Full Transcript

Michael Finocchiaro

And we're live. Welcome to ⁓ a Finos AI Across the Product Lifecycle podcast. My name is Michael Finocchiaro and I'm joined by ⁓ Michael Bogomolny. Did I get that right this time? And Masha Petrova. I've got sort of a Russian origin call today, which would be a lot of fun. They're founders for ⁓ Infinite Space and Nullspace respectively, which are pretty awesome firms. And I'm really excited to have you guys on.

Masha

But go more, will Come on.

Michael Bogomolny

All right.

Michael Finocchiaro

Are you guys going to introduce yourselves?

Michael Bogomolny

Marsha, please.

Masha

Oh, thank you. So hi, I'm Masha Petrova. It's great to be here. I'm the CEO and co-founder of Nullspace. We are a simulation software, electromagnetic simulation software company. And my background is, we were just talking about this, bachelor's in mechanical engineering and PhD was in aerospace and now I'm in electromagnetics.

Michael Finocchiaro

Awesome.

Michael Bogomolny

My name is Michael Bogomolny, I'm the founder and CEO of Infinite Form. Infinite Form is a software company which redefines design for performance and manufacturability. My background also mechanical structural engineering with PhD in structural optimization and post-doc in applied math. that's my background.

Michael Finocchiaro

So I guess you got lucky, you actually did something in your life that was close to what you in your career, because I haven't done any mechanical engineering ⁓ in a very long time. ⁓ So can you just, because Amasha mentioned that NullSpace is doing a simulation in electromagnetism, what is Infinite Form doing?

Michael Bogomolny

Infinite form actually helps engineer to design parts, which takes manufacturability and performance into account and generates design fully automatically.

Michael Finocchiaro

Awesome. Thanks a lot. So today's podcast is always on this particular show. I try to get through the hype on AI, right? To understand really how people are using AI. And we all had this moment in 2022, 23 when ChatGVT suddenly burst on the scene and everybody was freaking out that they could talk to the computer and get English sentences back. Were you guys... Bullish or skeptical when you first started playing with it? You were like, my God, this is going to be changing everything and it'll be ⁓ inside of every software I'm going to use for the next forever, basically. Or were you sort of like, I don't know, it's a little bit hokey and maybe a little fad. What do you guys think? You want to start, Michael? Yeah.

Masha

⁓ you go first. And by the way, can I just say in this podcast, we have Amasha and Russia and Michael is actually Misha. So you have a Misha.

Michael Finocchiaro

Two Mises, right? Two Mises and a Masha.

Masha

There's actually a now, very popular Russian cartoon, it's Masha and a Bear, cold, because it's another word for a bear. Anyway, sorry, not AI related, but...

Michael Finocchiaro

That's hilarious. It's funny because you're like, you're like mashing a Misha sandwich because you're like in the middle from in my screen. So there's some like some Michael, were you bullish or skeptical back when, you know, opening I broke broke the internet with chat GPT.

Masha

you

Michael Bogomolny

Well, historically, I can tell you that I've been looking at the space of machine learning, which LLM is an involvement of machine learning. But in general, it's the base algorithm was developed back in 60s, 70s, and now the implementation came. And it's basically you take historic data and you train or build a model based on it. And then every new request to get results. So in actually my. HD and postdoc I was involved in progress in problems that we solved using training, learning some regression models. haven't called them. machine learning back then we use them some surrogate modeling at different names now they called machine learning I quite well aware of the limitations these methods have in ⁓ engineering space and I was initially skeptical but they what these models are able to do in language space it become pretty impressive, I must say. So in the beginning, I was quite skeptical, but as we evolve and we see the progress which was done during the last few years, it's pretty impressive. ⁓ In 24, we have our philosophy around the usage of AI. Yeah, but that's the answer I can give you for your question historically, how it evolved.

Michael Finocchiaro

Yeah. Well, we're going to get into how you're using it with Infinite Form in a minute. Marsha, what about you? Were you like super, super stoked or were you also a little bit, we'll see.

Masha

No, I got to tell you, so I don't like technology. No, just kidding. I'm skeptical about technology, probably because I'm an engineer. And so I was actually when the whole AI hype was happening. First of all, I'm allergic to hype because my background is in marketing. So I understand how hype is created. I also come out of former Soviet Union. So I know what propaganda is and how that's created. And so that made me very, very skeptical. And so for the longest time, I avoided like

Michael Finocchiaro

Yeah.

Masha

doing anything to do, having anything to do with AI. And then it's actually an interesting story. Last Christmas around Christmas time, we were fundraising. So this was before we raised our first ⁓ venture capital round. And it was extremely stressful time. And I've had like a million things I had to do because I was ⁓ the salesperson for the company or building the company. You know, you're as you all know, you're building the plane as you're flying in it. ⁓ And I came to a point where I was just really depressed and I'm like, there was a lot of stuff happening in my personal life. I just too much going on with my company. And at that time, I have a couple of CEO friends that we are sort of in the chat group that we talk about different stuff, support CEO group. And I reached out to them and one guy who was a PhD actually from Stanford said, listen, why don't you just try AI to help you get like your thoughts organized and help with productivity? He's like, it helped me get my last SBIR grant and so now the company has money. So was like, okay, fine. If I helped you get money, I guess I'll try it. And so I tried it mainly for productivity. And what I found that it did for me, and it seems very simple, but it was huge at the time, was it helped me get through writer's blocks. So I'm dyslexic, I am not great at typing. And so for me to write an email takes a really long time.

Michael Finocchiaro

you Mmm. ⁓

Masha

because I have to check it a million times, I miss spelling errors and spell checkers don't really catch all of it and so on. And so I found just doing stream of consciousness like dump into AI ⁓ and then just having it process my emails all of a sudden like 10x my just productivity and then helping me kind of organize my thoughts, right? So I don't use it as a therapist. I still think it's a tool, you know? But just realizing, wow, you know.

Michael Finocchiaro

It was insane,

Masha

this is technology that could be very helpful on the productivity side of things. I don't rely on it to create things, so like create emails. I don't rely on it to, we don't use it in our company to like create original content, because you just get an AI slop. But to use it to help you like get through what I call writer's blocks or creativity blocks has been really helpful. And then we hired a head of marketing who is awesome. And she shall we say, a younger generation, so they're just, she's more... natural with new tools. So she was using AI. So we use Notion's AI system, which is awesome for our project management and things like that, for example. And she was just so fast. I'm like, I can't keep up. unless we all start using AI for certain things, it's just you use a human, you become so much faster. So that was kind of our foray into AI side of things.

Michael Finocchiaro

Yeah, it's been pretty astounding, especially on the latest ⁓ thing for Google. It's just been insane what you can do with it for that. So ⁓ in terms of using AI, ⁓ I'm imagining that the developers that you have at Infinite Formanel Space are using Cursor or Windsor for ⁓ a GitCub copilot or something. How have you managed that ⁓ going forward? I talked to one startup. He said, only let the senior developers use the AI stuff because I want the junior developers to learn how to program before they're using this assistant to write all the code for them, which is one idea. But I'm wondering, from you guys, how are you guys, you know, how are your products being developed with or without AI?

Michael Bogomolny

Well, so ⁓ we in company, we use AI in areas. First, inside the product, the interface between users request, understanding users request and setting it up in the right way in software. This is one way. So AI basically interfacing between the user and the engine we develop. Second is once result is done, AI is able to analyze result for users. Let's say you get design and can analyze its functionalities in terms of mechanical performance and specify critical areas, ⁓ areas of concern and recommend ways to improve the design because it's all access to data and language. So it does it pretty, pretty well and help to. redefine problem in more accurate way to improve the results using our engine. actually AI in our case is implemented inside the GUI, inside the product itself. In terms of development itself, ⁓ Again, language, code is a language and many of our developers are not restricting the opposite. And actually in startups early stage like ours, we need to move faster and restricting from junior developers not to use AI. Don't think it's a smart decision because you have to become productive. And you know. We're not the college, we're not in the university to teach. We hire people who highly educated and have expertise in certain way. But if you need to develop to develop some routines or some functions which do basic stuff, actually AI is very good in this. So it's shortening development cycles significantly. ⁓ Not in the first attempt you get perfect code, but with little debugging fairly quickly you can get result. Again it's not it you cannot tell build me code which optimizes parts but it's unable of doing it but if it is some routine that gets certain parameters and should perform certain task it's pretty good in this. So this is where we use AI in our product.

Michael Finocchiaro

Thanks. How about you, Marcia?

Masha

Yeah, I think very similar. we my I'm lucky to work with a brilliant co-founder who's our CTO as well. And so for a long time, he wasn't he wrote and worked on AI based optimization type projects that they're not part of null space yet, but they'll be released in the future. And but he himself was just coding it himself. But he's just super fast and really smart. And like he could stay on, you know, in in line with with how fast AI was developing, but at some point he's like, okay, we're going to use AI tools. So now we've got like every possible AI tool for development there is, I think. ⁓ And I think the philosophy of the company on the development side is let's try everything and because it's clearly going to be part of the future. And so similar to what Michael is saying, ⁓ how they're using it for development. We don't have many junior engineers at the moment, so I can't tell you what we're going to do once we start bringing them on board. We're also lucky in a way that we sort of cross over the physical and software side of things. We developed software for engineers who are creating physical things, whether they're planes or cars or, you know, defense systems, whatever. And so we have the luxury of our engineers and our developers have to understand how physical world works. And you kind of still have to know how physics works. And that's still not, you know, solvable by AI. But for the coding part, I think it definitely makes everyone a lot more productive on the software side of things. The other thing with Nullspace is because it's a sort of, we call it a modern simulation tool compared to some of our competitors that have been around for decades. It was written with, it's almost like, you know, my CTO, Daniel predicted that AI is going to be the future. So it's all written on up with a Python API as part of it. Oh, someone's joining. Really not. Python underlying Python API, which is the language of AI. So the idea is that null space can be driven by humans or it could be driven by AI. And so we're not quite there yet, but we're building out that the background, sort of the driver that's going to drive the simulation through Python.

Michael Finocchiaro

I don't think if you're Are you also finding that stuff like PRDs that you're just generating them directly rather than having someone actually to write down a PRD and stuff like that? It seems like that's another place where AI is being used a lot in development organizations nowadays. mean, for like those kinds of very text oriented tasks, it's like writing a PRD, writing a user story, ⁓ stuff like that. Are you seeing a lot of AI happening there, which is one of the things that's accelerating ⁓ the adoption and stuff?

Masha

We're experimenting with that, I would say, on our side.

Michael Finocchiaro

How about you, Mike?

Michael Bogomolny

Yeah, all the activities related again to language, to marketing is are done on our side with different AI tools. Additionally, we are aware that we started training our own AI model, which is smaller and more dedicated to our needs. Yeah, kind of, yes. And because, you know, relying on...

Michael Finocchiaro

an SLM? Okay.

Michael Bogomolny

calling the API, some external models can be quite cost prohibitive. And we found a way that if we train our own model, we can be more accurate and we can make it more cost effective.

Michael Finocchiaro

Gotcha. And I was just wondering too, like, when you're doing the more technical stuff, when you're doing the more algorithmic stuff on design or on simulation, is AI coming in useful there too? Maybe to read through a textbook and find the three or four formulas or maybe you're just like, you know it all by heart and you're like, I don't know, I know it's that formula and I'm gonna, I mean, I was just wondering how much AI is helping in this really, you you guys are both very deep into the. the math more so than maybe some other ⁓ people would be.

Michael Bogomolny

When our developers jump into new new new space of development, let's say new physics new formulation One way is to find books start reading another way is to call AI and gives you formulas and summarizes many things So when especially on exploration or literature survey Stage it helps tremendously to to focus and find the right information first second they Infinite form software, it's a combination of geometry, simulation, FEA, and also manufacturability. So it's multiple discipline which we have to work in good synchronization. So we need to find the right information, right data, tools for each discipline. Also AI helps a lot. However, I must say that... The engine itself that we developed, it's fully deterministic. It's numerical deterministic engine. One of the reasons are that you probably experienced when you ask AI exactly same questions, give you a different answer. And from our engineering ⁓ community very much concerned because they know if they have, if they draw shape and you give comment and give you the same shape, same shape, if you ask twice, three times.

Masha

Yeah.

Michael Bogomolny

And semi-simulation space, right? We run finite element simulation. If it's exactly same setup, you will get exactly same result. yeah, no, Maybe if you use iterative solver, convergence, different machines can be slightly different. Some tolerance can be, but in general, it's same. So if you ask, that's a concern of engineering community. I will ask the same question. I will get different results, how I can trust what I can trust. So the fact that that we have deterministic engine which solves problem and find exactly same solution many times you can ask ⁓ helps a lot. think combined, besides AI we have tremendous. evolution of hardware technologies as well. what GPU, modern GPU allows us doing nowadays, it's unbelievable. Of course, having such an advanced hardware ⁓ pushes in this developed better solvers as well. So, because if you look 15, 20 years ago, the solvers were completely different. Now, because they were CPU oriented, now GPU oriented solvers are completely different and we are able to achieve tremendous speed up. And our philosophy is that ⁓

Michael Finocchiaro

Of course.

Michael Bogomolny

everything you can solve numerically just go and solve don't call AI because you cannot rely as well as many of surrogate modeling techniques I want to be careful, not reliable because you don't know if you're looking into interpolation and not extrapolation. When extrapolation, it may be very inaccurate. Interpolation works very well. And, ⁓

Michael Finocchiaro

Right.

Michael Bogomolny

Yeah, I mean, also we want to be cost effective to train this machine learning model that takes tremendous amount of compute time. But if you can develop very fast solvers, which can solve problem directly and quickly within seconds, rather than training for days model and to obtain approximated result kind of questionable. So that's, that's my takeaway where in engineering world AI must be used for actual geometric implementations versus the text.

Michael Finocchiaro

Calculations. You have the same experience, Masha, or maybe slightly different?

Masha

Yeah, for sure. think it's, you know, for any of us who've been trained in engineering, especially if you have any kind of like graduate postgraduate degree, we have been, it has been drilled into us to avoid disasters, right? Like to avoid, minimize risk is kind of part of our nature. A lot of times, even for those of us who become entrepreneurs and like risk in some extent, we also have to make sure we're controlling other factors to minimize the risk, right? Because

Michael Finocchiaro

Yeah.

Masha

This things that engineers build, again, think about it, this is critical infrastructure. It's the bridge that your car is driving on. It's your actual car. It's the plane that you're flying in. It's the, you know, the satellite that you're relying on for communication. If stuff just starts breaking, it's very bad. And so engineers for generations have been kind of trained to make sure that stuff does not happen. And so simulation tools over the last, you know, 50 years or so since they've really kind of come to their own. engineers rely on simulation as part of the R &D cycle, right? Like it's very hard to find an R &D process in any decent size organization where simulation is not part of the process. so simulation has been proven out to, like Michael said, to give you reliable enough results that you can sort of rely on and minimize the number of actual physical prototypes that you're doing. ⁓ And so I think it just goes against every fiber of an engineer's being to all of a sudden start You know? Yeah, exactly. Start asking them. Especially, you know, just having your own human experience, like asking Chad to be your Claude anything or like last night I was trying to do something with my notion AI assistant. So I built myself a virtual assistant and notion notion is an awesome sort of productivity tool. They've they've built out some really cool stuff in the last six months or so. And I was screaming at it. I was cussing at it.

Michael Finocchiaro

probabilistic system.

Masha

Because it kept lying to me, right? Like you ask it one thing and it just flat out lies. And you're like, no, I put in your instructions. You did not make stuff up. Don't lie to me. Like I told you a million times. ⁓ yes, you're absolutely right. And it lies again. So as any engineer, if you have experience with ⁓ experience like that and go like, I don't want that anywhere near my product that I'm actually having to develop. And so I think until that. how that gap is bridged and how we do it is, you know, it's gonna be some time.

Michael Finocchiaro

I had this idea that, well, because we haven't even talked about agentic stuff yet, but if you're dealing with agents, I think any decision would have to be done by a committee of agents, and on that committee of agents, there should be a myminities agent. The agent's like, I don't know, I'm not gonna say how great you are, in fact, I don't believe anything you're saying, so prove it to me, otherwise, no way. I think you need a myminities agent because the whole thing of yours is the most beautiful, wonderful, awesome, smartest person ever, and. by the way, know, Charlie Chapin ⁓ just came out with a new album yesterday. that kind of thing. It's just crazy. ⁓ And ⁓ you mentioned GPUs. I'm just curious. And, know, this is unscripted. was just thinking you guys saw how Google with their TPU blew away everybody on the GPU, which prompted Nvidia to go off and invest two billion in synopsis to try to lock down chip design for them. Do you guys see?

Masha

Yeah.

Michael Finocchiaro

a battle coming up for between GPU and TPUs that change a lot how you guys are coding and therefore you might be thinking, wow, maybe I should go and look off at that Google stuff and maybe use a TPU instead. mean, is this something you're actually thinking about? Or you're like, no, it's just GPU and I'm not writing code that deep anyway, so it's not going to have any effect on me. I don't know. Just a question.

Michael Bogomolny

Yeah. Well, on GPU side, there is a good infrastructure with CUDA and all other libraries to develop. GPU, have to, I honestly, we have to explore it. don't have too much information about this. have to explore what, what requires to update the code or change the development because somewhat the infrastructural things are established and it's, it's working. I can't comment at the moment around GPU too much unless

Michael Finocchiaro

Yeah, of course. Right.

Michael Bogomolny

it's done in such a way that it's easy to code because hardware and software have to communicate. I think Nvidia have done great job with CUDA and all the interfacing.

Michael Finocchiaro

Right. And that's probably also what we need for quantum, right? Because quantum is so hard to code to, that'll be a stumbling block as well. I'm sorry, Marsh, I didn't let you comment on my TPU GPU.

Masha

Yeah, no, think we're, I mean, our development is definitely looking into it already. My CTO tends to stay on top of that kind of stuff. I'm not saying it's solved. I'm saying like, it's definitely on the radar. Nullspace actually has both GPU and CPU acceleration. So we work with both. And, you know, if you look at just moving away from the technology side and just, tend to be, I'm an extrovert, so I like people.

Michael Finocchiaro

wow.

Masha

So probably in the wrong field. No, just kidding. But I do enjoy people and so just following the leadership at say OpenAI versus say Nvidia and Google and how they structured the company ⁓ and just the interesting details about ⁓ Who is at the who is leading those companies and how they're leading it let's just say for me it's clearly that Google has this kind of like very systematic strategy that they've been following along. They're a very established company. They've been around for long time. There doesn't seem to be a lot of really crazy chaos. There's regular chaos like any large company, but nothing too egregious.

Michael Finocchiaro

Nothing like open AI firing the guy and bringing him back and all that kind of shit.

Masha

And all the other stuff. Yeah, exactly. ⁓ And so to me that signifies they're probably gonna be around for a while. So TPUs will probably, I mean if I had to guess, right? TPUs would probably be around for a while. ⁓ It's also, Nvidia is very, very, very heavily investing in simulation as I'm sure, you know, ⁓ at their developers conference when was it in April of this year. I mean, the head of ANSYS I think was up there and Cadence and Synopsys and everyone and their mom was at that.

Michael Finocchiaro

yeah.

Masha

conference from the simulation side, right? So it's interesting that Nvidia is clearly leading into engineering software space. But yeah, think it's, or guess this is as good as any.

Michael Bogomolny

Nowadays I can tell you almost every good workstation for engineering simulation have very healthy GPU inside and it can do a lot of nice things.

Michael Finocchiaro

⁓ yeah. one of these I bought my Mac M2 Pro because the Mac has just awesome GPUs on the Apple Silicon. Yeah, and that was really interesting. think, Michael, you already talked about how AI is integrated in your product. said you've already talked about that. But I don't think, Masha, you told us how a user interacts with AI when they're using null space from the user level or is it from the like the plumbing level?

Masha

Yeah, so Yeah, so not from the plumbing level. So we really are, you know, for a long time, we were really focused on our solver being pure physics solver. It's full fidelity solver. You know, we're solving the governing equations of electromagnetics. ⁓ And where the AI comes up is actually our next release that's coming out this January ⁓ is going to have some really cool AI things for CAD cleanup. So for simulation, you have this cycle, right? Where you have to build out your CAD and then you have to mesh it, right? In order to mesh it, ⁓ you have to clean up the CAD, especially for electromagnetic simulation. There is a specific process of cleaning that CAD. You want to remove a bunch of small parts that are not contributing to the electromagnetics part, but can make the simulation very lengthy, especially if you're running full fidelity simulation, right?

Michael Finocchiaro

Nice. Yep. Right.

Masha

And so for a long time, that was a pretty tedious process. And with current competitor tools, it's a relatively tedious process where you have to go back and forth with the mechanical engineers, removing this stuff in SOLIDWORKS because it's easier in SOLIDWORKS than in the actual ⁓ simulation tool. So we're coming out with a new release that allows you, it's got some really cool AI tools that allow you to remove and clean CAD quickly and help you mesh things. much faster as well. So that's our first sort of foray into the AI side of things. And then on the roadmap, we have AI optimization tools that will make that loop of CAD mesh, simulate, manually look at the results, determine, these results don't really quite fit to where I need them to fit. I go back and I redesign or I tweak something on the simulation, right? That's a pretty manual cycle right now, even though everyone again is saying, that it's not and they're adding AI to it and whatever. It's kind of like slap a chat bot on it and call it AI optimization. So we're not doing that. We actually have a very ⁓ robust AI optimization process in place that we are looking to commercialize in the next year, two years or so.

Michael Finocchiaro

Awesome. It was a nice.

Michael Bogomolny

If I may add, know, we talk about our engineering community and engineers used to work with certain software and don't like to change because they just get used to GUI and usually engineering software GUI usually have hundreds of buttons and it's hard to learn. And this one of the kind of areas which, which

Michael Finocchiaro

Yes. Hi.

Michael Bogomolny

I have, I think limits from engineers to evaluate and test and finding better tools for work. in infinite form software. It's, it's a software which combines simulation and manufacturability and geometry. So, and we try to build as intuitive and easy to use GUI and because users have to define what material am I going to use? What interfaces must be preserved? What loading conditions are applied, static vibration, dynamic load cases, what's design goals and how I plan to manufacture this part. What is going to be. the tool size from which direction I'm going to machine. So this sounds like pretty complex, but if you put the relevant text explaining exactly your intent of designing the part, AI actually does pretty good job because our AI tool understands our workflow and sees our options in GUI. So it really helps to set up. all the requirements and the software in a very effective way just from understanding of design intent. Yes, this is this just want to do.

Michael Finocchiaro

That's good.

Michael Bogomolny

Clarify how we are using it and as Masha said once design is ready again all the FEA results It's also text, know exactly your stress deformation everywhere in your model You know, what's the yield stress of material and you know if you were over? engineered to under engineer this material that this part and if there is a need to to to Optimize the part if you achieve some safety factor requirements. Plus, if you're an engineer... the company and you generate a design, have to go to a manager and present the work, right? So then have to sit down probably half a day, create like report. And AI it's a perfect tool to get all your results and summarize them in a nice way. Instructured report, this actually also implemented in InfiniForm software. Once it's ready, you're ready, you can ask to create report, summarize it and have it.

Michael Finocchiaro

Very cool. There was a nice comment from Peak Tech. It sounds like a company rather than a person, but it says, great topic. The bridge between simulation and generative design is exactly where the future of engineering lies. I think I can agree with that. I wanted to say, because the next question is around, if you guys were skeptical and bullish back then, how do you feel at the end of 2026? But it doesn't sound like, you guys have done much with agentic stuff. that's sort of the next wave is MCP, agent to agent kind of things. So I'm wondering, is that where you see your kinds of software going? Like where you could actually just use an MCP to grab the algorithm, do whatever it needs to do, and then pass that on to the next step and in some ways do workflows? Because let me back up for a second. Aorabari at PTC, talks about agency of agents and like you've got an advice, which is just tell me what I need to know and assist like, give me one thing. I'll let you do the one thing. You know, here's three options, choose one of the options and I'll do it for you. And then automate, is here's a workflow, whether it's a very like three steps or 10 steps. Here's a whole workflow where I'm going to trust the AI to do it. I think I'm hearing from you guys that automate, you're not there at all. You're like, maybe.

Michael Bogomolny

No, we exactly did.

Masha

Well, the problems are the same, right? Whether you call it an agent or MLM, the issue is still the same. You can't rely on it with certainty. sorry, Michael.

Michael Finocchiaro

No, no, no, I wanted to have the debate on it. wanted to just talk about, know, because you guys are both more on the skeptical side. as we move, as the rest of the market moves to this MCP thing, where do you think ⁓ we as engineers are gonna sit on it? So go ahead, sorry, Mike.

Michael Bogomolny

So, thank you. No, it's not. I think maybe it was misunderstanding. You're not on skeptical side. We just on the side where there is a reasonable, there's a good reason to use AI Android.

Masha

side of reality.

Michael Finocchiaro

Hahaha

Michael Bogomolny

Yeah, and we're not what you just described this exactly we to AI copilot or AI agent that we deployed in infinite form software. This is exactly how it works. You can come and say, I'm designing automotive part and the AI agent is asking, can potentially demo it. It's asking what material you're planning to use. would recommend these two or three types of material. You say, I would like to use this material and it's set up all the mechanical parameters and then it

Michael Finocchiaro

Okay. Okay.

Michael Bogomolny

asking what manufacturing technology would use. If it's photo motive, I would recommend either die casting or extrusion and you text. So it's really a conversation between a user and the agent. This is how it's implemented in InfiniiForm software at the moment.

Michael Finocchiaro

That's really cool. It's awesome. And Marsha?

Masha

Yeah, so I mean, we don't have any kind of agent type application at the moment. ⁓ And again, you know, if we do in the future, the issues are still, again, the same issues, right? So I think our philosophy is that I would agree with Michael. I don't think that we're skeptical. I think we just know the limitations of the software and we are just we can see through the hype, I guess. And, you know, we're just realistic. It's a tool. It's a great tool, it's very interesting, but it's not like, it's not God. Let's just say that, right? It's tool. so, as long as we keep that in mind and don't buy into the hype and just stay with the reality of what it can and cannot do, there's actually, for folks who are interested, there's a really great podcast that I just found. It's called ⁓ The Shell Game, and the guy is a journalist. And so he is experimenting with building a company. So he's been following AI for the last 25 years or something like that. And he is building a company with AI agent co-founders. And so like his employees are AI agents and he's documenting the whole thing in the podcast. It's really interesting because you can really see, you know, it's not an engineering podcast, but you could really see the limitations. We're going to run into the same limitations if we apply these agents and whether it's engineering or anything else.

Michael Finocchiaro

Nice. That's really cool. What was it called? The shell game. Nice.

Masha

And so again, because our customers are designing real physical things that can really hurt people, know, a plane falls out of the sky, or the antenna doesn't communicate the critical situation, it could really hurt human beings. I think we have to be just very, very careful with where and how we apply those. And again, simulation sits right in that space where

Michael Finocchiaro

Right.

Masha

engineers are used to relying at this point, relying on it to reduce physical tests and physical prototyping. And so, you know, to rip that out and start throwing technology that is, you know, like to hallucinate every once in a while or like, it's just, it's, not deterministic, just like as Michael mentioned.

Michael Finocchiaro

Well, then before we go into the customer discussion with how we end the call, I'm just thinking too, I think there's a lot of anxiety, right? There's a lot of these things on, know, AI is gonna steal my job. Since you guys are both founders and you're both engineers, you've probably already given thought to the question of the future of work in the sense of where, if you were a student and now like you were 15 years ago, would you have changed the way you were studying? Would you have changed your subject? Would you have done the exact same thing or would it? What kind of advice would you give to these younger people that are like, Jesus, what am going to do? And how am going to have a job? Because it'll be an AI guy doing my job. So I would like to put that to you guys as leaders. And what do you guys think? What was your advice to them?

Michael Bogomolny

You know, how do you know that AI lies? You apply your own intelligence, right? This is where it gives you incorrect result. So engineers have to, I may be old school, but you have to have very fundamental knowledge and understanding of physics. I'll give you an example. In 90s, 40s, 1950s, people built...

Michael Finocchiaro

Right.

Michael Bogomolny

buildings or bridges without computer. Why? Because they had understanding of fundamental physics and engineering. I still think in my generation, I my PhD in all my studies in Technion in Israel, we were kind of prohibited from using finite element software, at least in bachelor and master's to do all these calculations manually. You develop good feeling.

Michael Finocchiaro

Yeah, that's true.

Michael Bogomolny

of engineering and good understanding and it actually follows it like you learn the language it follows with you the entire life you have you look at the structure and you have understanding if it's good or not I think if you young generation and you see so many different tools you click the button it gives you result you have to have this good engineering judgment to understand

Michael Finocchiaro

The intuition,

Michael Bogomolny

exactly, but this intuition comes in fundamental knowledge. ⁓ I see many students, new generation coming with use of too much of tools and have less understanding of the fundamentals.

Masha

studying. And Dursun comes from a lot of studying.

Michael Bogomolny

And as we progress further into this AI world, I'm asking questions, give you the answer. You what, how do you know if it's the right answer? That's my biggest concern. Engineers have to still ⁓ keep, keep learning and knowing the fundamentals. And actually it's also, I would like to point it for schools to universities, colleges, when they teach, you must know fundamentals. before going ⁓ into using different tools because garbage in garbage out. You can ask wrong question. It will give you some answer. You will not know if it's the right answer. You're going to produce part and it can break. Of course, that's why Masha, as we say in many companies, if there's a critical components beyond simulation, there are also physical tests, but not always you can do physical tests. And you, some companies you rely on simulation or you rely on the digital tools. And if you rely on digital tools, they operate. these digital tools must be qualified people.

Michael Finocchiaro

Thanks. And Masha, you have the same advice or?

Masha

Yeah, so I completely agree with Michael, but I also have a soapbox. So I have a soapbox on that getting degree in engineering and by that I mean real engineering. don't mean software engineering. ouch. type of engineering. Yeah, I know. Sorry. Let me rephrase. Let me rephrase. Getting a degree in a physical type of engineering, an engineering that crosses both worlds, the software part of things.

Michael Finocchiaro

Oof. Half our audience just left. ⁓ No, it's okay.

Masha

an actual physical world where you can with your hands go and build something because for now we're still are living in a physical reality, right? And so ⁓ that will force you to do what Michael is saying is you will have to understand. Sure, you can use AI tools and whatever, and sure, you can like, you know, upload your homework into chat, GPT or whatever. But at the end, still will have to you have to build something that functions a certain way, you're going to have to learn it. And so, and also, you you're not gonna be replaced by AI, because you're actually building something with your hands in the real world. So if you don't want to go out there, and right now I hear there's actually a lot of recommendations for kids, because colleges are so expensive, is to go to a trade school. Go learn a trade, which I think is actually a great idea. Go, you know, become a welder or learn something with your hands. I think engineering is such a great ⁓ major to learn things that are physical, so you're not gonna be replaced by AI. But at the same time, you're using technology and lets you leverage that to kind of, you know, go into a more advanced technology fields. So I would say, yeah, good. So I was just gonna say, the world just needs more engineers. So if any young people are listening and wondering what major to choose, mechanical, electrical, aerospace, civil, great, great majors.

Michael Finocchiaro

Maybe, no go ahead, you would say, finish your sentence. Yes. Nuclear. Well, just maybe the thing I was just thinking was that when you're as high because you guys are CEOs, you're hiring people. like, what's the criteria on that? Like, are you looking at them being having the fundamentals, but also having already done their homework on knowing how to use AI? So they show up and they've already got ideas on how they're gonna do it or is it really just the fundamentals? Have you already started, has your thinking started turning that way in terms of being a hiring person?

Michael Bogomolny

Well, specifically in our case it's highly technical product

Michael Finocchiaro

Right.

Michael Bogomolny

So we have to have an expertise and we have many PhDs in this field. I would say it depends on the person, not on age. If person is open-minded and is ready to explore new things and is not afraid of experimenting and we encourage experimentation as well. Because you need to keep up on what's going on. And as we develop tools to make our customers more effective.

Michael Finocchiaro

Okay.

Michael Bogomolny

we also want to become ourselves more effective using some of these tools. ⁓ We don't put specific criteria. We just look at people who are open-minded, flexible, and this is super important in startup in such an early stage because things change all the time and some things don't work. You have to be creative in finding much more effective solutions. You think this gonna work and it's not working. You have to find the way to make it work and have to think out of the box. This is the type of people we look for.

Michael Finocchiaro

Same for you, Marsha.

Masha

I mean, yeah, yeah, exactly. And if you think about it like this, let's say you have two candidates interviewing for whatever position, right? And one, and let's say it's some sort of an engineering position, like application engineer, let's say. ⁓ And one is clearly highly sophisticated and really understands physics and engineering and clearly has proven that they have done that. But they say, you know, I've never used AI tools, just haven't really used them. But they're clearly smart and they are eager to learn and they're flexible.

Michael Bogomolny

Thank

Masha

but they're like never touched AI before in my life just for whatever reason. Code everything by hand just because I am fast enough. And then you have a second candidate who is technically not great, but loves AI and says, I use all these tools like all the time. Like which one would you hire? Do you know what mean? Like the one who could pick up the AI tools and learn. I mean my 94 year old grandfather, he found out about ChatGPT and like talks to it all the time now. He learned how to use it, right? Or.

Michael Finocchiaro

Ha

Masha

someone who is not really technically skilled but is excited about AI and use all these AI tools. That's easy to pick up if you're flexible and fast. And so I think right now, just look at what is like a harder thing to achieve. Is that an engineering degree or is it talking to chat GPT for a year? Like what's harder, you know? And pick the harder one and that will sort of... ⁓

Michael Finocchiaro

⁓ So let's switch gears and talk a bit about ⁓ AI when it hits the real world. I think of AI, when I think of companies, I think of their digital maturity, because you've got to have a certain level of maturity to have AI in your organization. And so I think of organizations that started at one, which is like email and Excel still, right? They're barely using anything that's digital up to... the mythical company that's at five having autonomous, agentic, ⁓ adaptive digital twins. mean, basically nobody's at five, almost nobody's at four. ⁓ First of all, so first of all, question is how do you gauge it? Maybe it's industry by industry. So I'll just throw in that I talked to a guy, you've probably seen him on LinkedIn, Peter from XM Pro. ⁓ He has a, it's more of an OT system, but he's looking at... mining and he said you wouldn't believe it but mining like 10 years ago was like in the stone ages and now it's probably the most advanced industry in the world because ⁓ autonomous vehicles to go down in the mines find stuff and bring it back up and all these vehicles have to use electricity because if you're using diesel you basically kill everybody inside the mine and so you can imagine a grade like this going in an electric car the battery is going to die so you've got to be like and i was like holy cow that's amazing right so you guys are doing space and you're just doing aerospace and space tech and Where do you see the industry in terms of digital maturity at, you without naming any names, but just, you know, in general, what would you say that would be between the one and five?

Michael Bogomolny

I would say everybody in space defense aerospace industry understand they have to become more efficient ⁓ How you become more efficient You need new tools if you continue using saying it's it's delicate situation menu of blue chip customers multi-billion dollar Revenue companies they have built workflows around the existing tools for decades And it's, it's a safe path. However, the current workflow is very slow. If they want to move faster and more effectively, you have to use more advanced tools. And I think what we see there is realization from many of these big companies, they must. they must start adapting more advanced tools because they will increase productivity by orders of magnitude, not 10, 20%. It's really orders of magnitude. One of the kind of again, limitation when we talk to space, defense, aerospace is that due to ITA regulations and different regulations, they cannot use cloud-based tools. And... most of AI tools embedded into cloud-based tools. this is, I would say for some of these companies. ⁓ obstacle in some sense because you know, IT or cybersecurity always step in and say, what do do with AI and how we implement AI? But we have, have to be isolated because there are regulations, they have to be isolated. Now, some of them use GovCloud, private clouds, and there are some solutions. However, it's not easy. And I think these big companies have to become more and more flexible. Smaller companies, some companies who who are less restricted, more flexible in adopting those tools. And I was surprised some of customers who would never believe would use cloud product, they're using cloud product. ⁓ Overall, I think the industry have to do some also shift. in understanding that new tools must be adapted. Not necessarily we're gonna replace the entire workflow, but there are some tools like infinite form. We're sitting on top of existing workflow, not displacing any CAD simulation or manufacturing software. We just bring in layer of intelligence, which will understand all the requests and will... generate in minutes optimized results, which will satisfy all the requirements alongside of the workflow. And yeah, that's.

Michael Finocchiaro

But between one and five, so you're saying in A and D they're probably between two and three, something like that. And they're aiming at four, basically, that's what I'm hearing. Would you say that's better? Yeah.

Michael Bogomolny

It depends on the industry again, some industries and also size of the business, small medium size companies adopting those tools much faster. But still I'm surprised, but many of the huge enterprises started adopting these type of tools. I would say even three,

Michael Finocchiaro

Right. Okay. Thanks. Okay, really?

Masha

You would say three, four? I would say two, three probably. So we were born out of defense. So we kind of natively understand the needs of the defense space. And so yes, cloud is not a thing still in defense and aerospace companies, right? So we're all on-prem, although internally we run no space on AWS for ⁓ any kind of internal testing or customer problems or something like that. I think to add, I completely agree with what Michael said. And I think

Michael Finocchiaro

Yeah.

Masha

If we zoom out a little bit, we're just in the beginning of this transformation for the engineering industry specifically. So consumer industries obviously were much faster on sort of engaging the whole AI experience. The critical acquisitions that happened, at least on the simulation side of things, also with Altium on the ECAD side last year, right? crazy number of acquisitions. Not only that, but also if you look on the startup side of things, mean, Michael is one of them. How many startups in engineering software actually raised huge? Yeah. Yeah, huge number. That's right. You did a post about that, right? Which is not typical. mean, when I was just started fundraising, when we started fundraising and just talking to investors only three years ago, investors didn't know what the space was. I literally would get comments like,

Michael Finocchiaro

390. I found 390. That's just insane.

Masha

What? Like this seems really boring and slow moving, like simulation. What? Like tools for engineers. What? Like what is why we, right? And now all of a sudden look at how much money is floating around and how many. Is it X, Luminary Cloud? There's a bunch of them that are coming out. And so I think what's important and at the same time with the acquisitions of Ansys and Altair ⁓ and you know, now the huge investment in Synopsys.

Michael Finocchiaro

look at physics X, right? A billion dollar valuation is.

Masha

the engineers are kind of looking around, at least in our space, going like, wait, hold on, we've relied on these tools for 40 years. We didn't have to look for anything else. I mean, think about it, for four decades, there was nothing else you needed to look at. You had Ansys, had DeSoe, you had Altair, like, PTC, and that's pretty much it. And all of a sudden, they're like nervous because there's no alternatives. And it's not like the field is just overflowing. mean, on electromagnetic simulation side,

Michael Finocchiaro

You That was it.

Masha

We hardly know any startups in the space. so, okay, but the point being is, to Michael's point, is these companies are, this is like year two, let's say, for these larger companies to start looking in the new technologies and also how fast AI is evolving, right? So it's like a combination of things with these huge acquisitions happening all in one year with these big.

Michael Finocchiaro

I think I found one, but we'll talk about that later.

Masha

pushes from OpenAI and Nvidia and so on happening also in the last three years or so. ⁓ It's very early. But I think like Michael said, know, the successful companies will figure out a way to work with engineering workflow because engineering workflow is so precious. I worked at a startup six, seven years ago acquired by Ansys now where the CEO came in and his first mandate was to take on-prem simulation software that was developed and break it all and force customers on the cloud. Like basically said, we're not selling any of our premises. You have to buy cloud license, right? That which is just idiotic and did not work and you know, whatever. And then they were sold to Kansas. you can't do that because these methodologies that are so ingrained in engineering workflows are so ingrained that I think like Michael said, if you can come in.

Michael Finocchiaro

No. ⁓

Masha

like NoSpace as well. So we're not coming in and trying to break anything. We fit right into the engineering workflow. Yes, you have to learn the new interface and there's a little bit of a learning curve involved, but that's it. Like there's no, you don't have to reshuffle everything else. Licensing structure. We just, I won't say the name, but there's a very large defense company that we just closed a deal with. Just to freaking get them the license and get the invoice, you know, was... like a complete nightmare. We had to talk to like 20 different people just on the invoicing and, you know, send the links of the license side of things. And that is, I mean, imagine if that happens for invoicing, that happens all throughout the engineering process. And like, if you try and come in there and go like, oh, use our AI, like, let's scrap your entire processes. There's thousands of people who are part of these processes. can just do that. So, yeah.

Michael Finocchiaro

issues. Of course. ⁓ Yeah, so like last question before we do closing remarks, but last question. So ⁓ you said that you guys feel that the companies are relatively mature, but ⁓ when you put in ⁓ infinite form or in all spaces, there situations where there's sort of an aha moment, like, wow, if I'd done something a little different, if I'd broken the data silos between departments and be able to move that data between them better, in terms of data governance ⁓ or between different jobs because you're, I mean, Michael, you're working on the boundary between engineering and manufacturing and Masha, you're between design and engineering and manufacturing also. ⁓ And those tend to be in the past, we're all siloed, right? There's different people didn't even talk to each other. ⁓ When you guys put in your solutions, is there an aha moment or a ripple effect where people like, my God, these guys have to talk to each other as we can't go faster and then so forth. So I'm just wondering like, is adopting Bleeding Edge, awesome software that you guys do. Is that an absolute ⁓ win in terms of moving your needle towards the right, towards more digital maturity and having a better collaboration between engineers?

Michael Bogomolny

That's great question. That's exactly one of the problems. talk about engineering workflows currently. If you look to design to manufacturing workflows, there is design department, there's simulation department, there's manufacturing department, each one using different software packages. The communication is going via emails or there's a project and this project manager when he got a location of engineers from different departments. And again, this back and forth. And that's why design of simple part can take weeks today and big companies because of this. And one of the reasons, not only Siloed workflow, also about the flow of the data in between departments.

Michael Finocchiaro

Mmm. Right.

Michael Bogomolny

And this is what InfiInform is breaking, trying to break and we're breaking it currently. Because before, one of the problems also, designer designs parts, throw it over the wall and saying to manufacturer, go manufacture it. What if designer would have all the manufacturing... parameters in early design stage and would incorporate them in early stage, it will eliminate ton of iterations. It will make manufacturing much more cost effective and straightforward. And actually, if you look into mass production, these are millions, not tens of millions of dollars can be saved for the customer. And this is exactly what we at Info.Info recognized. We need to bring this layer of intelligence into existing workflows where the data from design and simulation and manufacturing department

Michael Finocchiaro

Yeah.

Michael Bogomolny

will get into the project in early design stage to be incorporated in the early design stage. So we're shortening design cycles by order of magnitude and closing this disconnect gap in between different departments. So that's exactly you pointed out the biggest challenge in the industry and how we're it.

Michael Finocchiaro

Very cool. Is that your observation as well, Marsha?

Masha

Yeah, I think we're trying to be very delicate about how we come in, right? So luckily for us, engineers are used to having multiple simulation tools because they like to play with them and because no simulation tool is ideal yet. ⁓ They like to have multiple simulation tools. So for us, that's already an established situation that we're coming into. ⁓ But we are seeing, we've actually had this with a customer where the customer had a very convoluted workflow set up. because they were working with one of the ⁓ current tools that is competitor to us, the same place that Michael used to work at. And that tool couldn't handle these large, complicated simulations. So they had to break it into pieces and build this. This was for some radar scattering type of application. Break it into pieces and then run simulations separately. And then they had this very sophisticated workflow that they built out over the years. They were very proud of it. And Allspace can come in and actually solve the entire problem. Like you don't need the workflow. And we see this like almost like a stupor in the engineer's eyes where they're like, but we build this whole thing to accommodate this other thing. Right? And we're like, well, we don't need it. And it's like, well, how do you answer that? Right? Because, but you're to break all of this. And now with the insecurity that's being introduced by AI about everyone's jobs, right? Michael Finocchiaro (1:00:12) Ha But ours goes to 11, but it goes to 11. Masha (1:00:37) I think engineers are going to feel more more insecure about what they do and probably will become more and more protective of their workflow, which makes sense. Although hopefully if you're a progressive engineer, you'll just kind of embrace the tools and try to figure out, okay, well, how can I become better and kind of move along with these tools? But that's not always the case. And so we ran into that already. you know, unfortunately that's where the world is going. So in that case, we go to the engineering manager and... show them like here's what you're doing now here's what you could be doing and it's it's obvious and so they will have to override the old workflows but it's gonna be a process and it's gonna take time I think it's not gonna happen overnight although you know VCs and the the tech hype space will want you to think it'll all happen tomorrow but you know it's gonna take some time Michael Finocchiaro (1:01:25) I think we're looking at two or three years at least before. One of my podcasts, asked, I think it was Sebastian Voekel, I was asking like, it a bubble? And he said, no, no, it's not a bubble, it's a forest fire. Like it's gonna burn out all the crap and the firmly rooted companies will survive and all that garbage of, ⁓ we'll see, it's gonna be an interesting couple of years I think. ⁓ Masha (1:01:50) like that. That was great. Michael Finocchiaro (1:01:53) I've had a great time. really awesome talking to you guys. thank you for being so open and ⁓ insightful ⁓ in terms of AI. And any closing remarks before we end the session? Michael Bogomolny (1:02:07) I would say it's exciting time and an exciting space. And I think you have seen lately, even Jeff Bezos invested in the new project. So there is understanding there have been also the previous attempts to revolutionize the manufacturing space. And... Michael Finocchiaro (1:02:18) ⁓ yeah, 6.2 billion in Prometheus. Michael Bogomolny (1:02:33) I think there is an understanding that should be unified solution and platform to connect all pieces together because the way it has been really silent and AI can be this trigger, which will collect many of broken pieces together. However, I do feel new, new design tools needed because the current, the, the, the existing tools cannot allow Efficiency as everybody expected and we see that in order to manufacture parts effectively design should be Straightforward and taking into account these manufacturing constraints and currently everything is disconnected. I think we live in this Exciting time when some pieces can come together and puzzle can be and can come together and and workflows can be fully automated ⁓ And yeah, Infinite Form is playing significant role in this space. We found exactly the gap in the workflow we developing and actually our official launch is going to be in January, second half of January next year, where we going to launch our platform. And by the way, Masha, as you said, we have on-prem solution for defense aerospace enterprise companies, but we also have cloud. Michael Finocchiaro (1:03:46) Nice. Michael Bogomolny (1:04:01) offering for all the rest and we're super excited Michael Finocchiaro (1:04:06) I hope it will do it. Masha (1:04:07) Where are you launching? Sorry, Michael, are you launching an event or a trade show or anything like that? Michael Bogomolny (1:04:12) It will be before the event. will be also, by the way, Michael, you asked what events next year going to IMTS in Chicago, we're to attend and we also going to be in 3D experience world as we partner for the SO system, SolidWorks and we have few more events next year, but this is a major. Michael Finocchiaro (1:04:18) Yeah. All right. Okay. Awesome. How about you, Marcia? Masha (1:04:34) Yeah, and we're launching our newest Nullspace release is coming out in January as well. That's the one that has the exciting CAD cleanup features. We're going to launch that AIAA SciTech in January in Florida. And we're going to be at IMS and Phase Array Symposium as well later this year. Michael Bogomolny (1:04:41) Thanks. Michael Finocchiaro (1:04:43) Nice. Michael Bogomolny (1:04:48) Okay. Michael Finocchiaro (1:04:50) very cool. Awesome. Well, I hope maybe we'll see you at a couple other ones that we've talked about. yeah, and maybe we could get another Misha's and Masha hour maybe after your launches and see how it goes. That'd be fun. But someone's got to bring the vodka and somebody else says to bring the samovar and then we'll be all set, right? ⁓ Michael Bogomolny (1:04:58) Congrats. Masha (1:05:02) It'd be great. Thanks so much for putting this together. This was super fun. You Michael Bogomolny (1:05:13) Let's do it. Masha (1:05:14) That would be awesome. There you go. Michael Finocchiaro (1:05:22) Thank you very much. Thanks to the audience for joining and we'll see you on the next AI Across the Product Lifecycle. It's been great. Thank you. Masha (1:05:29) Thank you, this was awesome.