Entrepreneur Balint Pasztor: How DiffuseDrive Disrupts the $124B Data Market

During the rush to scale AI, innovators, entrepreneurs, and engineers are identifying the bottlenecks that hinder progress. Easy access to quality data is emerging across various industries, fueling a timely search for solutions.

Balint Pasztor is the co-founder and CEO of DiffuseDrive, a next-generation physical AI company transforming how Fortune 500 companies access real-world, high-volume data with unprecedented speed. 

Within a year of launching the physical AI startup, the company is disrupting a $124 billion market with a platform already adopted by Fortune 500 companies. It converts generative AI into high-throughput, synthetic data generation pipelines, replacing data chaos with clarity and data scarcity with abundance in days rather than years, providing safety-critical synthetic data for camera-first autonomy, robotics, and defense systems. 

Balint offers a compelling story of how innovative entrepreneurs seize opportunities when they see firsthand the technology challenges faced by large corporations.  

Hear how he and his co-founder put in their sweat equity and leveraged their relationships to move past a minimum viable product to end-user Fortune 500 clients, gaining timely funding to scale their startup. 

Transcript

Jeanne Gray: I am Jean Gray, publisher of American Entrepreneurship Today and host of the podcast series Experience Voices, where I talk with highly accomplished people who share the critical elements that led to their success.

Welcome to the Experienced Voices podcast. Today we are joined by a visionary 

Ballant Pastore, the co-founder and CEO of Diffuse Drive, who is solving one of the biggest bottlenecks in the AI world, data scarcity. Diffuse Drive is a next generation physical AI company that is fundamentally transforming how Fortune 500 companies access the massive volumes of real life, high fidelity data they need.

Ballant and his co-founder are disrupting $125 billion market, and they're doing it with unprecedented speed and scale. 

Thanks for being a guest Balint. 

Balint Pasztor: Thank you so much for having me. 

Jeanne Gray: Well, it's a pleasure to have you on experienced voices because of what I've learned about your disruptive technology and its impact it could have on the scalability of ai and so timely of what is going on. So let's start by you defining or explaining to listeners what your breakthrough technology is.

Balint Pasztor: Absolutely. Most importantly, it's not just ai, but physical AI is what we are tackling, something that touches the real world. And usually what I bring as an example is when you have to teach a system or train a system to interact with the real world, of course you need. Experiences, or more specifically, the way that humans are, taught in from the very beginning by our mothers, is to show us, good and bad examples of what we can do. Let's say you see a dog for the first time. Your mother teaches you that it's a dog. And then again and again, they don't describe it to you as it has four legs. It has a body, it has a head, but rather it just points out that it's a dog. And the same way we are training machine learning systems is we are showing examples.

And these examples could be anything. And of course we are talking about robotics and we need training and validation data for, robots. So one thing is, train them and the other one is to test them against a golden set of data. And then of course, usually how it goes is it comes from the real world.

So people are collecting data, but by the nature of the real world, we are limited by it, especially on one hand we don't have the scale. The other hand is of course, we don't have , the means to collect everything. And also not everything happens when we want it to happen. So just an easy example.

What I usually bring is that everyone drives a car or most people drive a car. And whenever you drive and you're following a car in the, traffic, If you were to replace your eyes with cameras that are recording, you are recording a lot of data. But the problem there is that you are seeing essentially one thing.

The scenario doesn't change. The, the car in front of you doesn't change. And of course, when you're. Recording all these you are interested in the most diverse data that you can possibly get. So this is what we are generating with, with Diffuse Drive. Instead of getting it from the real world, we are generating all the synthetic or rather AI generated data.

Not just cars, but also drones, also robots. So all of the different domains of robotics is falling under us or physical ai. That's what we are doing. In a nutshell, 

Jeanne Gray: well specifically describe data scarcity and why your solution is addressing that challenge that large companies are having with their engineering teams.

Balint Pasztor: Absolutely. it's a finicky question 'cause we have data scarcity, but it's not that what people think. So we have a lot of data in the word. The problem is again, what you are recording or what you are getting from the real world is on one hand very limited because you are seeing the same thing.

So the usefulness or the impactful data is very scarce within that. Within the already existing data. The second is there are all those edge cases, long tail cases that are actually making or breaking the algorithms that are really scarce. So they don't happen as frequently. You might not be able to get it and what.

Let's say I'm, gonna bring another automotive example and then hopefully we'll stay away from that for some time, is that Tesla is one of those companies that has figured out how to collect masses of data in the real world. And of course they are San Francisco Bay Area or Silicon Valley company.

And they are recording a lot of stuff around here, especially California is really good and big on Teslas. And what you're seeing is the same thing driving over and over and again over the same course of track. You're gonna be really good at that. But of course, outside of it or out of distribution is really not.

And the same thing for all the. robotics and diverse. So the data scarcity is where , the algorithms break, not the general cases. And I was reading an article a couple days ago and where we are right now in the robotics industry is.

Pretty much a hundred years behind for everything that is written. So we have language models that are being developed on a daily basis by these companies like OpenAI, and they are digesting all the information that we have. Written as, humans even going back to a hundred thousand years ago with the wall paintings.

Unfortunately for robotics, we are literally in the very first age ages or very first decades of collecting data. And of course we want to get to there faster than a hundred thousand years. But might as well just supplement it with synthetic data or synthetic means. And that's why we are doing diffuse drive.

Jeanne Gray: Well, let's start or go back to the origin of your, idea is you are in a situation or in the industry where this exposed you to this problem. Share a little bit about your background and, and where the light bulb went off to pursue this type of solution. 

Balint Pasztor: Absolutely. I mean, working at Bosch, the German automotive supplier and working on their au autonomous driving stack , we face this very problem ourselves and my co-founder, who used to be my right hand, my technical lead he also worked in the very different places within the autonomous driving and the autonomous driving stack that Bosch had before.

And we just realized that it's, a problem for everyone. I was also selling autonomous driving related parts to Tesla, Arabia and Lucid and all these different ones in the US already. And we realized that it's a problem for us. So not having enough data, it's a problem for everyone.

And hence we started the company tackling the driving part. But we realized quickly that it's not just driving. It is data scarcity for many, many different endeavors. regarding in the air autonomy so drones or aerial vehicles. On the surface of the waters. So unmet surface vehicles and everything.

And we realized that it's not just the civil applications that we have tackled before, but also defense applications. 'cause at the end of the day where robotics are being applied. First at least, is where it is dangerous, where it is unsafe for people. So essentially replacing crude or un crude systems or robots that are operating at the very edge of the technology or the very edge of our capabilities.

Jeanne Gray: This was a big step for you to take. You're working for Bosch, but you've got some great background. You found a co-founder. So what was the first step that you took to implement or execute on your vision? Did you establish an MVPA minimum viable product, or did you have access to systems that could allow you to do some experimentation?

What was the very first steps to get out of the gate, so to say? 

Balint Pasztor: Absolutely. I think that the very best way to do it is just to do it right. So of course we started with Roland, my co-founder, who I am really thankful to be working with. 'cause we essentially compliment each other on many different fronts.

And more specifically, he's the builder. I'm the seller. Or I'm, rather on the business side. And that's what we did. We built an MVP, a very early MVP even before having a company, even before having anything. And we validated with customers. I was part of Entrepreneur first in London. , Before we started the company.

So I am very thankful of on learning how to validate stuff in the wild or rather in the market and just apply the knowledge, right? We went to our customers, or previously my customers, and then hit up people in the industry through our connections and validated The problems they were facing.

, What are the things that they needed when it comes to synthetic data or the data scarcity? And then we realized that it's bigger than what we thought. So we started building it, and we built the first, very first one. We showed that to potential customers. They were already excited to have that.

And then we got our first, investments even before a company, because people also realized that we needed to build this. 

Jeanne Gray: When you, again, looking for additional funding, realizing that , you really had something and you could see the opportunity to scale, did you go through the traditional steps of putting together a pitch deck and.

Coming up with , a list of initial investors that you would approach. So share a little bit about that learning curve. 

Balint Pasztor: The way that we started in the first place is that, of course we wanted to go through these accelerator programs. I have seen the. Value that they are providing to early stage people, especially for those that are not as fortunate to grow up in the Bay Area or being exposed to more venture backed companies in the first place.

Especially coming from a more humble beginning from central Eastern Europe. I would say that the difference mainly is that the mindset. So we wanted to put ourselves into a place where the mindset is, prevalent. And then while applying for all these we got already almost dragged into which in itself would entail that we didn't have a control over it.

But we let ourselves to be dragged into like a semi fundraising mode. And that's what resulted starting the company with already having investments in. 

Jeanne Gray: When you were pitching the first investors, how did you define for them your market share? Were there already competitors presenting something as innovative as startup or you were trying to get first to market so you could dis plant some legacy technology that was inefficient the industries?

Balint Pasztor: There were a couple indicators, right? There were already companies operating in the market doing synthetic data or doing simulations. And our whole pitch was built on, yeah, simulation don't work. Here is what we are doing. That's one thing. And the other thing is people were just starting out GPTs of the world.

So OpenAI has just released ChatGPT a little bit sooner than. Or earlier than we started the company and then those were the, pinpoints where everyone was realizing that we are running out of data that we can use for training , all of our ai or at least the, real world data.

So we needed to get some more from the real world and the combination of all these our first investors realized it, but also we realized it that we are on the right track. And as much as would love to see that it. It was more serendipitous. I would say things fall in place, of course in venture or in everything in life, but in venture as well, you need to count your luck and you have to expose yourself to be lucky.

And I think we were also in the very best position in this era to, start a company like this. 

Jeanne Gray: Did you have to change any of your assumptions of when you, did your MVP your prototyping where what you learned out of the first trials were very insightful and you said, I or your team, I need to adapt to what we're doing because a lot of entrepreneurs marry into that.

The solution at the get go is the only way to go. Is adaptability important in your early stage of developing the technology? 

Balint Pasztor: No, absolutely. I mean you have to listen to what, what people want. And of course y Combinator is always saying build something that people want. They cannot be more right on that.

Especially that you have to just learn about your, customers. 'cause even if you are building the very best technology you are not necessarily building the very best product. That people are using. And at the end of the day, this is the product company. And our bread bread and butter comes from, deploying a cutting edge technology that, is being brought over from research into development. So our, whole system is being built on top of the very last.

Technology that comes out of the industry. 

Jeanne Gray: So when you brought your technology to an actual end using customer can you share a little bit about the satisfaction of sort of the aha moment of the client where they've confirmed to you? You you've got something. 

Balint Pasztor: It's so funny 'cause whenever we are, pitching, we always bring them into like, okay, let's play a game.

Can you guess which one is real and which one is generated? And the reason we are bringing this is of course in a visual. Domain where we are operating you have to see the results yourself. And as a human, we are really good at pattern matching and especially pattern finding and seeing where the things don't add up.

And in this case, hallucination or just failures where the imagery falls apart and this is what I'm doing with them. And then seeing is believing, right? So we are just showing them how good the imagery is and how indistinguishable it is from real life. And we started calling it Real World Great Data, it looks like, feels like, speaks like the real world, but it doesn't come from the real world. And that's what we are always anchoring around. And then of course the aha moment for them is when it's like, okay, yes this is something that looks like my. Own imager or my own camera, my own hardware, my robot, what it sees from the real world, except what there is in the imagery or in the in the data that we are generating is something that they have never seen before.

Jeanne Gray: So you're bringing enormous quality, and I'm hearing that you're also bringing cost savings to them compared to the way that they were previously managing and establishing their data. So is that your model based upon not just the quality, but you can go to a customer and they can see how implementing your technology will save them a lot of time and money?

Balint Pasztor: Absolutely. we are saving time and money to them, but there is one caveat that I always want to bring to them since we are giving them essentially an access to data that they would otherwise have to take years to collect.

We are giving them a time machine and they are sparing money. Or rather resources later down the road that they don't have to use and don't have to utilize to just linger around with like , semi coverage of their data or proposed data sets. So that's something that I really like to emphasize to them is that we don't sacrifice on quality.

We don't sacrifice on quantity and diversity because we don't want them to sacrifice on time. 

Jeanne Gray: So how are you running with this? So you have your, first adopters, you've validated the model, you've gotten some seed or pre-seed money. So what was the next phase that you entered to? Was it to multiply the number of early adopters?

 How did you sort of accelerate, , and build momentum behind your early success? 

Balint Pasztor: So, of course , there is a lot to do. At an early stage company we raised $5 million altogether to build out and scale our business. And most importantly, where we are heading is twofold. One is of course we have already deployed it in civil applications, which is very important.

Now we are deploying it at the defense applications. So for that, we are setting up our, staff for success at the federal level as well. And at the same time, we are going towards, into all things robotics with our SAP served product later down the road. So right now what we are doing is of course in the early ages of the industry where we are seeing the robotics data being proliferated, and we want to be the people that are in the forefront of it.

Jeanne Gray: Do your early adopters or your testimonial clients, are they acting as the entree to additional doors and additional clients? A lot of aspiring entrepreneurs don't really appreciate the value, how the first few customers can change the direction of a startup. 

Balint Pasztor: Oh, right, right, right. Absolutely. Most of our early customers are definitely among those that are the people that are just advocating us and especially coming from there. And with their backing and with our, with our resources, they are also being very well. Compensated for that in terms of like, hopefully we are showing them just as much appreciation , as they are showing to us.

And at the same time we are expanding. So our customer base has multiplied over the last couple of months, years. So definitely, definitely being grateful for those early customers. 

Jeanne Gray: So where did you deploy your $5 million in funding? What was your, say, top two or three areas or priorities?

Balint Pasztor: I mean, of course we are, we are hiring the best of the best. We have European engineers, we have American operations. And more importantly, right now we are building up our American engineering as well. So that's mainly the top of mind. And of course for us competition at the end of the day where we are , still being squeezed on, or as everyone else is, having enough compute. But at the same time we are running our most efficient models for most of our customers so that we, don't have to sacrifice on quality. 

Jeanne Gray: Well, you've touched upon different industries that you're entering. Step back a little bit and share.

How you formulated your marketing approach to , your next stage. Did you bring on a CMO did you bring on a marketing consultant? did you write a business plan and execute , the marketing part of it? 'cause , as we're talking, it sounds like you're moving very quickly, but then there's.

Pieces that you're putting into place that's allowing you to reach that success. And I imagine having a good marketing strategy is part of this puzzle that you've been putting together? 

Balint Pasztor: I think we built a really good team over the last two years of the operation of the company that is allowing me and the whole company , to thrive in the marketing part.

Personally, I like being involved in the marketing plans and like. Even down to the creating side. So being able to just formulate our thoughts and show it to the word. I think that's the easiest way to put it into marketing. But at the same time, like marketing involves a lot of different things.

How the market is changing, where we are tackling it, and stuff like that. So. I think that's something that doesn't leave the C level for a really long time, if ever, if I'm being honest. Even at bigger companies. And that's the same for us. It's like the leadership still takes care of that. Of course marketing implementation.

That's another, question, but for sure what we are saying, how we are saying it, and essentially how we are. Talking about ourselves as a company , that's something that is still in, in the realms of what we are doing. 

Jeanne Gray: Well, product market fit. I guess if you could clarify a little bit further you've shared a lot about how you.

Introduced , your innovation to clients that you had prior experience with and they validated your model. but I guess maybe impressing you a little bit on the whole aspect of creating the message into all of the other areas of defense and robotics that is there an online strategy?

Is it one of personal networking. Are you leveraging what the accelerator contacts that you had, because that's where I'm seeing some common questions pop up with those who are bringing innovative technology to the market is where do they budget for it and how do they get that initial momentum that all of a sudden.

, they have gravity to their marketing efforts it's easier to accomplish what they're doing versus stalling and saying, I've got something really good, but the doors are not opening. 

Balint Pasztor: It is all the above. I feel like you have to stand on multiple legs and especially like you have to put in the sweat equity for most of the stuff.

Of course you have to network, but I don't like calling it networking 'cause that means that it's highly transactional. You have to build relationships. We are in the people to people business. And you have to make others. Not just think, but also believe in you. And especially understand that you are not gonna go away.

You are here to solve their problem. And then also you have to make them understand that you understand their problems. And of course having the problem set itself for ourselves in the very beginning. as like our own very much helps. And since then, the building in defense, building in, robotics yes, we just have to listen to our customers, how they are formulating their problems, more specifically because the value proposition is not exactly the same for everyone.

For some it is finding the really last edge data or the edge cases or long-term cases You have to find the very first bits of data and start from there. And then everything in between.

For robotics, you might have a lot of data, but you don't have everything that you need for your needs. It's like somewhere in between. So I would say. Having the ability to synthesize it and then tell your customers what you are bringing to their table is always appreciated.

Jeanne Gray: Going back to your co-founder, so what hearing is you have the business savvy to. Address a lot of the scalability issues as far as , the sales and the marketing side. And he brings the technical and the engineering part of it. Is that where your, complimentary skills got you through the first, you know, couple of years because you had so much of the core skills between the two of you.

Balint Pasztor: Oh, of course. Especially talking about like someone that , is complimenting you from multiple angles. Yes. I mean, that got us through many things, starting the company together. We already knew what the other was capable of. And especially I could lean on to him. , On most of the things that I was not doing, which is of course, the technology.

He was the builder at the previous venture that we were involved in, and of course now he's the builder, he's leading the engineering team and, guiding them , into really good direction. So I really appreciate that part. And of course he doesn't have to do the things that I'm doing.

So Fact that we can trust each other to this level is just unavoidable , in a venture like this. And then, of course, highly appreciated. So 

Jeanne Gray: where do you see the company in the next couple of years? Is it a multiple of the number of employees? Is it a change in your location or, 'cause I believe you're working out of a couple of different locations.

how do you. Reset your vision. Now that you've got the funding, you've got the testimony, the early adopters. How did you revise your goals given what now, what you have in place? 

Balint Pasztor: We are building, building, building, building. That's, the main and most important part of the company is, of course we are hiring always we get the very best of the engineers on board with us and they are building the platform, the technology itself.

So there are a couple stuff Especially at a technology startup, it's always engineering heavy. And at the same time we are deploying in multiple different locations and regions in the world, in the United States, in North America, in Europe, in APAC region. So all of these require a certain level of presence.

And all of these presence is gonna be supported by our staff. But at the same time, we are building a self-serve engine or a self-serve platform. At the very end of the day, we are giving an access to our customers to generate their own data, and that's what we want to support for them which does and does not require a huge amount of engineering depending on how you're looking at it.

Jeanne Gray: Well, I very much appreciate your coming on Balint and sharing this technology that I can see is so timely to the market. So, I guess my last question to you is that especially since you know, you've said you've had prior venture experience and how well you and your co-founder have moved so far, so fast.

. Share with entrepreneurs what you feel a common trap is, or mistake that they make when they're launching, especially a new technology, where's their thinking? That is a part of a learning curve that they need to get past. what can you share on that? 

Balint Pasztor: I mean, for me, the most important one was just to do it right, like how Nike says it.

Just do it. But most importantly, you have to experiment. You have to be able to fail. You have to be willing to fail and just do it and try things. 'Cause something that's gonna stick will propel you far. But until you find it. You just have to try things, and especially in the very early ages, it , might be that you even have to pivot your whole idea.

We were fortunate enough , to not do that, although you could argue that in a lot of cases, the changes that we implemented can be viewed as pivot. We really think that we are solving the very same problem as why we started the company in the first place, but it might be a little bit different and it might be a little bit different tomorrow or the day after because we are more than happy to adopt it.

We marry. A problem set but not necessarily one problem or one technology. But the company is called Diffuse Drive. But if tomorrow a better technology comes out that we will be able to leverage, then it might be a different technology behind that. But we firmly believe that synthetically generated or AI generated data will solve the robotics problem.

And then we are bringing the best solution for that. 

Jeanne Gray: Sounds like you're saying to aspiring entrepreneurs, take the risk, get into the arena and then go from there. 

Balint Pasztor: Absolutely. 

Jeanne Gray: Well, it was a pleasure speaking with you Balint, and I look forward to hearing of your progress over the next couple of years.

'cause I know you're moving so quickly. So again, it was a pleasure. 

Balint Pasztor: We are doing our best. Thank you so much.

Jeanne Gray: You have been listening to the podcast series, experienced Voices. To hear more and subscribe, visit american entrepreneurship.com/podcast. Where you will also find a form for listener feedback.