How does 500 petabytes make ADAS smarter?

0:00 Introduction

Today in The Garage, we’re recording live at AutoTech 2026 in Novi, Michigan in the Detroit metro area. As one thinks about advanced technologies in vehicles, it’s impossible not to think about ADAS and autonomous driving. And we’re seeing incredible advances in those technologies all the time with things like lane keeping and, active cruise control becoming more and more commonplace, active emergency braking being required in more and more geographies in the future. And then, of course, it’s big brother autonomous driving where we’re seeing robotaxis with literally no driver. I know I see them driving near my office all the time.

To talk about that, we wanted to bring in experts. And my guest today is Nimrod Brickman. He’s Vice President of Business Development with Mobileye, an industry leader in ADAS and autonomous technologies.

We really cover a wide range of topics and about how their technologies have advanced in the past years and how they’re bringing these capabilities to more and more vehicles all the time.

Let’s go!

Welcome to The Garage. I’m John Heinlein, Chief Marketing Officer with Sonatus. We’re here at Auto Tech 2026 in Novi, Michigan. Nimrod, Welcome to the podcast.

Hey. You said it right. Hey. Thanks, John. Thanks for having me. My pleasure.

And look, we’re thrilled to have you here. Start by introducing yourself and telling us about you.

1:33 Meet Nimrod Brickman

Sure. So my name is Nimrod Brickman I’m the VP of Business Development in Mobileye. So I’ve been leading the business development in Mobileye for the past three years. I’ve been with the company for the past nine years.

A bit about my background maybe, so I’ve I came from basically investment banking, working with technology companies. So I’ve worked with technology Israeli technology companies trying to bring capital from different places around the world, focusing on China.

So I’ve been working with China for many many years. And then I started working with some automotive companies back then, and I got really intrigued.

Really interesting technologies, interesting companies coming, trying, you know, to swim in this very, very intense industry, And this is what got me trained into automotive overall. Yeah. That’s great.

2:21 International experiences in China and Japan

So many amazing technology companies coming out of Israel. Yeah. Yeah. It’s exciting to see the journey. You have to start by telling us a fun fact about you.

Okay. So going on with this China path. Right? So the first deployment from Mobileye, basically, my first relocation position was to China.

Oh, wow.

Yeah. I was relocated to China. It was twenty seventeen, twenty eighteen, pre COVID. I’ve lived there for a couple of years.

I’ve actually studied Mandarin back back from high from high school and then through university. It was quite useful. Which city were you in? I lived in Beijing.

I had some time pre-Mobileye in Shanghai as well. But I’ve lived in Beijing for three years. Amazing experience. Such a different culture.

Amazing.

That must have been such an exciting experience. My fun fact, I always share a fun fact back to my guests.

I did an assignment in Japan, in Tokyo. Also amazing.

Quite a few, about twenty years ago actually, which was incredibly rewarding. I like you, you’d studied Chinese. I had studied Japanese not in school, but I had been studying Japanese with a lot…we did a lot of work with Japanese partners…at Transmeta. This was several companies ago at Transmeta. And we had a number of Japanese partners and I’ve been studying Japanese quite intently and in fact, I’ve been doing a lot of business development. So an opportunity came up to do a very senior role in our Japan office and I was excited to do it.

Amazing. Also very difficult language, Japanese.

Well, we can have a whole conversation about that. Japanese and Chinese are differently difficult.

I00 percent. They’re they’re both interesting in their own respects and they have different and unique they’re actually, the difficulties between the two are different.

3:54 Mobileye’s History and Achievements

Very interesting. Very interesting. So that’s thank you for sharing that. Now tell us for our listeners who may not know about Mobileye. Mobileye is incredibly successful company and probably very well known. But for those listeners who might not know, could you tell us about the company and a bit more about your role? Sure.

So let’s start first with the company. So Mobileye has been around for the past twenty five years. We’ve basically, the company introduced computer vision for advanced driver assistance systems starting in 1999 with first projects coming into play in 2007 and onwards. Basically, it’s been a a world leader in the computer vision industry. And I think that just to give you some taste, by now we have more than two hundred and thirty million EyeQ chips sold.

Two hundred and thirty million.

Over that. Yeah.

That’s impressive.

It means that more than two hundred and thirty million vehicles are equipped with our our system.

Roughly one in every every eighth vehicle is using our technology, we’re very proud of. And now we we have multiple engagements with OEMs, which are customers, companies that we’re working with. The company has grown throughout the past few years, moving from a tier-two tech provider into a more of an holistic role, delivering a full solution of level two plus, level three, level four. We’re basically one of the companies one of the only companies who were working from base ADAS, meaning the very basic technology up to level 2+, level three and even to autonomous mobility. So this is with regards to the company itself.

That’s great. I think we’re going to get into all of those topics. And your role specifically is?

5:29 Nimrod’s Role at Mobileye

So I’m leading the business development activities. So basically everything which is pre-nominated, pre-sourced, I’m in charge of all the advanced development activities and engagements with global OEMs and just managing the relations and trying to get more deeper connections with our OEM partners and expand our business. This is basically the role.

And you’re based in the headquarters, is it Tel Aviv?

It is in Jerusalem. So I’m based in Israel, but we have multiple locations around the globe with multiple representatives. We have in the States, in Detroit here, and, and Paris, and Munich, Shanghai, Korea, Japan. So we have teams across the globe. Basically, in every hub that there is a big, OEM or a big part of the industry.

6:12 Evolution of ADAS Technologies

So you’ve been with Mobileye for nearly decade. ADAS has really changed a lot since you’ve joined. Can you tell us about that evolution? What are you seeing?

So it’s quite amazing because when I started back then in China, you know, it’s it was all about meeting regulation, front facing camera business, very basic compared to what we see right now, what we’re seeing right now. And it’s a significant leapfrog jump that we’re seeing within the industry. Basically, what the OEMs and the consumer are expecting from their system to deliver is completely different. Right?

If we’re looking at, you know, all of the level two plus, level three capabilities that we’re seeing out there, hands off, eyes off solutions, a decade ago that was just made almost science fiction. It was very imaginary. Right now, it’s it’s actually how OEMs are looking at their roadmap. And throughout your engagement, it’s a very interesting role that we’re we’re engaging right now.

7:08 Keynote Insights on Industry Shifts

You just finished a keynote just a short time ago here at AutoTech talking about Mobileye.

I mean, can you share with us the journey you’re you’re on and some of the things you shared in your keynote?

Sure. So first, I think it’s very much linked to this shift that we’re seeing right now and why the industry and the OEMs are taking this direction of of bringing higher levels of autonomy into the market. I think the interesting part here was how we’re taking different components that were used to be separated and segregated into the whole centralization approach. Right?

There’s the whole SDV approach and the whole centralization approach, bringing more components into the same one. So within the keynote, I’ve scanned through the different technologies that we have, parking, driving, driving monitoring system capabilities that we are integrating altogether into the same system. And basically, it means that we’re trying to balance in between three different, pillars within the OEMs challenge. One is to bring a very high, performance of of the function itself.

Two is to bring it in a very fast time to market. And three, for for it to be very cost effective and efficient. And this is how I address, you know, the key pillars that Mobileye delivers in this market into an actual product.

8:23 Evolving from Tier-2 to Tier-1 Supplier

Maybe additional fact here is when the industry is looking then at those technologies, they’re focusing, like, five to ten years ago, they were looking at, is it possible to deliver those technologies?

Right now, they’re looking at how you can do it at scale. Yeah.

And this is what I’ve addressed in our in our And and you mentioned earlier that you’d historically had been more of as a tier two supplier providing a component into another subsystem, but more and more you’re emerging as pretty much a tier one supplier providing a more complete subsystem or in fact multiple subsystems.

Is that a good way to think about your evolution?

Correct. An interesting point here is just that we’re we’ve gained a lot through our developments around level two plus, level three, and level four, which helped us in creating this kind of a holistic full turnkey solution around driving. So basically, if OEMs looked at us as a key technology partner for one of their components, they’re now looking at us as a holistic supplier for their advanced solutions.

9:22 Understanding Levels of Autonomy

You mentioned L2+/L3. I think that’s probably the biggest journey that the biggest hurdle that the industry is is looking at.

As the industry is looking to get to that higher level of autonomy and and perhaps you want to start by explaining for our listeners those those those levels mean. Hundred percent. What do you see as the biggest technical challenges to achieving those higher levels?

So first, you you pointed out a very good point. Right? Let’s start by explaining what what does it mean level two plus and level three. So level two plus and level three, the main differentiation is basically who is responsible for driving.

So level two plus, level two plus plus, there’s a lot of abbreviations around that. So the driver is always responsible. The driver is always need to be attentive and supervise the system itself even though if it’s hands free, or you can let the system drive, you need to be responsible for it. But level three is when the system itself takes over in certain ODDs, operational design domains that the vehicle needs to operate in.

Now this is a significant leapfrog jump in terms of technology that you need to deliver. First and foremost, you need to have much more sensors, Right? Much more compute in order to allow that to happen. And basically, what we’re trying to to change in between level two plus and level three, it’s not the driving function.

It’s the mean time between failure. Right. We’re calling it in the industry the MTBF. Sure. So the mean time between failure needs to be much higher, so it will allow the system to take over for the specific driving task. Right.

I think that’s a really interesting way to think about it because most vehicles today, the majority of vehicles that have lane keeping, and I know I’m not gonna name any specific systems, but many of these systems are generally l two plus because the expectation is at a relatively short amount of time within a few seconds, the driver needs to be able to take control if let’s say the system either doesn’t understand what to do or is an emergency comes up. With a a level three, I’m assuming that there needs to be now much more extended period of time that the system needs to be reliably able to handle things.

Correct. So you need to, to detect for much longer, and also you need to take into consideration what will happen if there is some sort of failure in one of the systems. Right. This is the way that we approach the level three as most of the industry is you have a main system and in addition to that, have a fail operation system. So it’s very it’s a very delicate topic and it’s a very safety oriented, discussion, not only the performance itself, safety first always.

Of course. And I I think so much excitement is around, ADAS and autonomous driving. It’s obviously, it’s a spectrum. And so much advancements in the past few years, it’s been incredible to see how fast the pace of innovation is going.

12:02 AI Integration in Autonomous Driving

In general, obviously, we’re seeing so much AI coming into vehicles, not the least of which, of course, is autonomous. As you’re bringing AI into there, understanding that, you know, AI systems are fundamentally nondeterministic and that’s their strength.

But it also can be a technical challenge. What are the things that you see as opportunity and challenges to bring in AI, and making them, as you just said a second ago, safe and and secure and reliable in this mission-critical environment?

So I think there is two ways to look at this AI revolution that we’re seeing ahead of us.

One is conceptual and one is practical. The conceptual part is whether you’re taking a a single monolithic approach of end to end AI. Basically, we’re calling it in the industry from pixel to torque or from pixel to control. When you have one single brain, which is a bit less on the safety oriented approach according to how we see it.

And the other approach to that would be a compound AI approach, which is how Mobileye is looking at it, and I’ll double click on that in a moment. The other aspect of it is very, very much not non conceptual, but practical. How do you take huge language models, visual language models, and visual language action components and put them in an online system, which also supposed to meet these both safety criterias and cost criterias of an automotive system. Right. So mobilized approach to that is is using a compound AI approach, which is basically a fusion in between both worlds. We’re basically relying on the safety element on the core long learnings that we’ve had for the past twenty five years with the composable algorithms, which are getting always better and better, introducing more and more components into the these systems.

And in addition to that, we have a frontier aspect of innovation in in end to end and large language models and visual language models aspects that’s currently where we’re integrating them together. Right? So we’re in a compound AI approach, we’re just merging in between the two and enjoying both worlds.

Should we think about that as you’re more fusing the models or that the models each have a specific function that then they collaborate? What’s the best way to think about that?

So we can think about it that the main semantic or human like driving will be done as tasks by or tasked by the end to end model itself. Right? So the more human natural type of behavior. However, you would still want the guardrails or the the boundaries of the system to be maybe a bit more rule based because there is the there’s always the long tail problem that you would wanna recover from. So basically, this is this is the approach. We’re just mixing in between them.

I think it’s very very clever.

So shall I think about that as kind of AI is watching over other AI?

It’s like there’s some deliberate separation so that there’s someone kind of looking over your shoulder or in a manner? You can think about it as yet another envelope, safety envelope on top of your core brain that will help to make sure that what you’re doing is the right thing to do.

And in a way that you could look back on it and track it back. Yeah. And not just, you know, it’s not like an open ended loop.

That’s right. We’ve done some work with safety certification. A year ago, we announced we’d been ASIL-D certified for a safety monitor for one of our products, which a very similar idea was that the safety monitor was able to look over the actions of one of our of one of our products before it took an action and say that it is or isn’t functionally safe in that situation. So it’s it allows you to not have to, in our case, functionally safety certify the entire product.

But you certify the the safety monitor. It’s a common approach in the industry. Correct. Sounds like it’s a somewhat similar approach.

Very much. It’s very much so that you are homologating the safety oriented system and not the whole system as is.

15:53 Mobileye’s 500PB Driving Database

So it’s well known that Mobileye has one of the largest driving databases in the world. I hope you can tell us a little bit about that. But because you’ve been in operation for so long, how does that database give you advantages and what are some of the ways you use that database to help your train and teach your your algorithms?

So basically, this is one of the core assets of Mobileye that we’re proud of. We’ve been around for twenty five years, as I mentioned, and we have different types of, of datasets. We have more than, five hundred petabytes of data that we’re we’ve aggregated throughout the world. It is, combined through, computer vision data, driving behavior data, and also RAM, which is a Mobileye road experience management.

This is a sparse information data that we’re harvesting on a crowd source base. And we’re utilizing this in two to three main ways. One, this huge dataset is helping us to train our end to end networks, our algorithms, and make them better. And we have real world data and use cases to to tackle with using this dataset.

This is one. Other one is that we have a huge dataset to validate what we’re developing. This is yet another very powerful tool. Other companies in the industry are working on simulators because they do not have this access to that much of data that we have as Mobileye has.

A third point is is how do we tackle the long tail problem. Right now, when we’re talking about end to end models, we’re seeing the long tail problem of edge cases. And right now using this vast amount of data that we have, we can be exposed to many, many different use cases and train our data to extreme scenarios. And we can basically improve our system dramatically.

It’s so important to point that out because I think if you if you think about maybe if you as a driver, if you think about your own experience, if you could imagine sort of mentally recording your driving, like ninety five percent of it’s utterly boring. A hundred percent. Maybe it’s ninety eight percent. But there’s two percent that’s totally terrifying or super important or safety critical. And that opportunity gives you the ability to find those two percent times millions of trips Exactly. Accelerate the study of those corner cases.

So right now, we’ve also this is a great point because right now, in order to to tackle this long tail problem, we’ve also just recently announced the CEO and the CTO of the company, Professor Amnon Shashua and professor Professor Shai Shalev-Shwartz, just published a new article about two new tools that we have in order to find those exact use case cases within the this vast amount of data that we have. One tool is Meteor, one one additional tool is Generio. These are two different tools that we’re using to basically using visual language models to extract those unique cases that are reproducible and can really create a significant change in the safety cases of those long tail cases.

Wonderful. And, you know, as of course, as ADAS has evolved, one of the things that you’re seeing is in certain geographies, certainly in the US and in in Europe with, you know, NHTSA standards in US and the the equivalent standard in Europe. You’re seeing now certain classes of ADAS features like automatic emergency braking notably and some others that are becoming standard in 2027 and beyond, for example. So how are you helping to bring your capability, which of course can be very powerful at the high end down to wider price points as that’s becoming more standardized across the line?

19:25 Bringing ADAS to Wider Markets

So I think this is a very good point when we’re looking at systems like I’ve just talked about today, the surround ADAS. It’s basically the idea is to bring multiple components and bring them into a single compute unit. The surround ADAS that I’ve mentioned a moment ago is based on EyeQ6 High. This is the brain of Mobileye. EyeQ6 High basically allows us to bring different components of driving, parking, visualization, and integrate them. So basically, if you think about it, an OEM would previously would have to spend hundreds and of hundreds of dollars on different segregated compute units that now can be integrated into a single one and deliver much more with much less.

So with with integration, the cost can come down. You still have high end solutions for some of the higher autonomous driving levels. But for lower end cars that just need the basic compliance requirements, you can meet that with different cost solutions.

So we can do that and we can even upgrade low trim vehicles to additional additional higher levels of autonomy with lower cost. So just take as an example Volkswagen. This is a partner that we’ve also announced a partnership with a couple of years ago that is taking the surround data center market. It is not well known for it to be a premium brand, but it’s taking still taking this ADAS level two plus into a lot of his vehicle brands.

That’s great. Yeah. L2+, you know, as you said, that’s sort of below the line where the driver is still responsible. But one of the things that you often see in L2+ is lane keeping.

It’s sometimes adaptive cruise control depending on the vehicle. And once you’ve used a lane keeping and adaptive cruise control, it really takes the burden out of, let’s say, highway driving. And, you know, in a traffic jam, for example, I’ve you know, in the Bay Area where we live, there’s tons of traffic jams. And I used to [think] “oh, it’s a traffic jam!”

Well, now, of course, I’m still I still have to pay attention, but the drudgery of a traffic jam is almost eliminated because the car does ninety percent of the work and you’re looking for, oh, do I have to take over? It’s really very, very pleasant.

I think if you’ve experienced this function, like going on a highway and just you don’t need to worry about the driving itself. Obviously, you need to supervise it.

But you can relax your body and just just ensure that the ride is going smooth and even even potentially also active lane changes. Right? That is also something which is very useful and until you experience that, you don’t really understand what it is. But once you experience it, it’s a significant change.

Active lane change, and my vehicle has that and I love that feature. And as a listener who maybe who hasn’t had that, it might seem gimmicky, but actually it’s really not. Because one of the things that’s a very frequent source of accidents is lane changes and hitting a car in your blind spot. I know my vehicle when I do a a lane change when I’m in lane keeping and and adaptive cruise control, it will wait until there’s a safe spot.

In fact, a matter of fact, has a cool animation. It says, I’ve found a safe spot. I’m going over to there. But it has made sure that there’s no vehicle in the blind spot.

So it’s actually more safe, not less safe, and it’s not a gimmick. It’s actually a very powerful feature that I love very much.

Think about it that you have additional eyes and additional sensors on top of the two that you have. Yeah. And it basically helps you to create a three hundred and sixty degrees environmental model around the vehicle and helps you basically to be more to have a more pleasant driving, but much, much more safe.

22:54 Conclusion and Future Outlook

Yeah. And and it’s so easy to think, “oh, I’m a better driver.” But the reality is every time I’m in adaptive cruise control lane keeping, the car is much safer than what I’m driving. And I’m big enough to admit that now, that’s why I use it every chance I get. This has been such an enjoyable conversation. Congratulations on all the incredible success you’ve had and on your recent keynote here. We wish you the best of success in the future, and and thank you for bringing your technology to bear in in more and more vehicles all the time.

Thanks for having me, John. Thanks.

If you like what you’re seeing, please like and subscribe to see more episodes like it both here at AutoTech and our other shows around the world. We look forward to seeing you in another episode of The Garage very soon.