In the race to commercialize self-driving cars, software expertise is a necessity. The CEO of a Portland, Oregon startup tells why.
By Mark Shortt
Of all the development programs currently being pursued by automakers and tech companies, perhaps the most radical are those aimed at bringing autonomous vehicles, or self-driving cars, to market. It might not be too much of an exaggeration to say that software will be to tomorrow’s cars what DNA is to human beings. For OEMs, it’s a race that cannot be won without the benefit of software that enables them to quickly and reliably prototype, test, and manufacture safe self-driving vehicles.
One company that could help them get there is PolySync, a Portland, Oregon startup that is developing a platform designed to help build the software that will run future cars. Likening its platform to “Android or iOS for self-driving cars,” PolySync aims to make autonomy system development feel more like mobile app development, freeing developers to focus on algorithms while spending less time on assembling the back end. Its platform is said to help simplify and accelerate the development, testing, and implementation of autonomous driving technologies.
PolySync (www.polysync.io) won multiple industry-wide honors in 2016, taking home the top prize in the Top Ten Automotive Startups Competition™ at the LA Auto Show’s AutoMobility LA™ in November, after being crowned Newcomer of the Year at the TU-Automotive Awards in June. Honda, Tata Motors, NVIDIA, Denso, and Dura Automotive Systems are among the firm’s customers.
“We believe that our job is to give automakers a platform that allows them and helps them to collaborate and to build an ecosystem around cars, in spite of the fact that their current tool chain and their current cultures make that very difficult to do,” said Josh Hartung, co-founder and CEO of PolySync, in an interview with D2P. “We’re really hoping that by showing automakers the value of our platform, and by helping them understand what safety looks like in the future world, we can convince them to do what really is best for them, which is to utilize open platforms and horizontal scaling to decrease their costs and their time to market.”
Hartung, a graduate of the University of Idaho with a degree in mechanical engineering and design, was chief technical officer at AutonomouStuff, LLC—a supplier of systems, equipment, and services for the autonomous vehicles market—before becoming CEO of PolySync. As head of a leading edge software company active in the automotive industry, he is among a new breed of engineers skilled at working in both hardware and software.
Design-2-Part caught up with Hartung recently to hear his thoughts on the new automotive ecosystem, why PolySync won the Top Ten Automotive Startups competition, and how he sees the industry going forward. Following is an edited transcript of our conversation.
D2P: Could you describe in general terms what PolySync is building, and how it works?
Josh Hartung: PolySync is building an underlying platform for building autonomous vehicle software. Actually, I would extend that. In fact, we believe that our platform will be used for building all of the software that runs a future car, and we call that the software-defined car. And practically, what PolySync does is a very simple thing, which is that it decouples the hardware from the software. It does that through a concept called abstraction. Basically, what that means is that it’s a standardized interface for interacting with different types of underlying hardware.
I can tell you it’s not the app developer that’s dealing with all that complexity; it’s the OS (operating system) itself. And so that’s essentially abstraction. The OS is saying ‘Okay, I’ve got all these different hardware components that I need to talk to. How do I provide a standard way of talking to them?
D2P: What motivated you to start PolySync?
JH: At AutonomouStuff, my job as CTO was to build up our services group. I went out to Tier Ones and OEMs, across not only automotive, but military, mining, et cetera, and, in that capacity, I got to see how everybody across the industry was building these cars, and the software for these cars. And across the industry, I saw that a lot of work was being put into, essentially, things that don’t make the car actually drive itself. Essentially, everybody was duplicating the same work around infrastructure. They were building the back end to their software, over and over and over again.
The real meat of the problem, though, and the very interesting work, was happening at the application layer—the applications that do machine intelligence, task planning and control, and things like that.
So we said, “What if we could take that infrastructure load off the OEM, or the development team, and allow them to focus on the applications?” That’s kind of what PolySync is: an underlying infrastructure for making it easier and faster to build autonomous cars.
D2P: What do you see as PolySync’s role in this new automotive ecosystem?
JH: We see ourselves as an underlying technology enabler of different approaches to autonomous driving and vehicle automation. Autonomous driving computers, and the technology that goes into that, are some of the most powerful that have ever been shipped on a vehicle. And so we see ourselves as an enabler of utilizing that very exotic technology more effectively, both in terms of development resources—making it easier and faster to build the software—but also making it easier and faster to update and to maintain the software.
Our mission is to build a future of software-defined vehicles that is simultaneously extremely safe, but also extremely flexible. The net result of having this type of infrastructure is the same change that you saw in the early days of PCs when going to PCs that run on Windows. There was a consolidation of one part of the problem, which was the operating system, and there was an explosion of different companies and technology that came out of that. So suddenly, you had all of these peripheral makers, and all these different PC makers, and all these different application developers, and an entire industry was effectively born out of that existence of a platform. We see a similar kind of “Cambrian” explosion, if you will, happening in automotive around our platform.
D2P: Congratulations, by the way, on winning the Top Ten Automotive Startups competition at the LA Auto Show. Can you tell us a little bit about how the entrants were evaluated and what it was like to be named the winner?
JH: I think they had, like, 50 or 100 competitors in that contest, all of them having some links to automotive. We simply pitched our vision and our product and our traction. We went up against larger companies, like Argus Cybersecurity, and smaller ones, like Evercharge, and I think we came out victorious because we have one of those business models that addresses what is simultaneously a really big acknowledged gap in the industry and one that is very, very hard [to bridge]. And so I think that’s the kind of thing that resonates with audiences, and with the automotive industry itself—big bets with big rewards if they pay off.
D2P: What particular strengths does PolySync offer from a software perspective?
JH: Our team is composed of folks from some of the biggest companies around. Our CTO came from Visa, where he was building a globally distributed payment platform. We have individuals from Intel and from Apple. We have some extremely talented individuals here that are specifically talented in building infrastructure, and that’s very different from a lot of companies in this space because most of them are searching for the next machine intelligence researcher or AI researcher. We search for people that have built massive, rescalable, and safe distributed systems.
D2P: Why is that particularly important?
JH: It’s similar to how you need to build a solid foundation under your home, to withstand what may come. You lay the walls so that they’re solid and rigid and safe, so that the house that you build on top of that will be there for a long time. That’s the same kind of approach that we take. We think of ourselves as foundational technology. Some of the major problems that are unsolved are “How do we make these autonomous cars safe? How do we test them? How do we validate them? How do we deal with the massive amounts of data?” So we think about building—how do we build a solid foundation underneath the exotic technologies that all these car companies are committed to supporting in the next five to 10 years?
D2P: In your collaborations so far with automotive companies, how have you found the interaction and communication, given the cultural differences between tech companies and automotive firms?
JH: We have very purposefully developed the ability to talk to both sides in our company. We believe that we can bridge the gap between Silicon Valley and Detroit, and we do so by creating a common layer of technology across which they both can communicate—not only in a practical way, using software, but also from a vocabulary perspective. We’ve done a lot of work here in trying to build a community mindset and an ecosystem that both sides can benefit from.
I would say that maybe three or four years ago, there was resistance to this type of narrative. But now, automakers are increasingly seeing the benefits of outsourcing technology. There needs to be a middle ground for us to get these products to market, and for us to really make a difference in the life-saving technology that makes the world a better place. There needs to be some type of consensus there, and we believe that we bring that to the table.
D2P: PolySync has been described as making autonomous system development more like building mobile apps, but you mentioned earlier that developers had been spending a lot of time on the back end. Why were they putting so much effort into doing that?
JH: Well, the problem is that these are very complex systems; they use a lot of bandwidth, from a network perspective, and they produce a tremendous amount of data. To give you an idea, cars will log anywhere from about 1 to 5 terabytes an hour. You’re doing that from, sometimes, 20 to 30 different sensors, and you’re doing it out of order, asynchronously. All of this data is coming in; it’s being produced at different rates and at different points in time. And so it’s a tremendously difficult problem just from an infrastructure perspective. But it’s not unsolvable, and it’s not even that hard once you’ve got it figured out.
But the problem is that getting to that point is very expensive. It takes specialized people a lot of time to figure out that solution, and at the end of the day, no consumer pays more for a car because it has a better underlying network, right? And that’s one of the challenges for auto, is that they have really ended up verticalizing in a way that has meant that they get some sort of short-term differentiation from developing technology internally, but they haven’t been very good at disseminating that technology out across the industry to bring the cost down.
Typically, in other industries, some third party steps in and makes something better than any individual company can do, and because they sell it across the entire industry, it costs less. But in automotive, they end up making everything themselves, from electronics to engines to suspension, to seats. They either design it themselves and have contract manufacturers make it, or they design and build it in house.
At the end of the day, what consumers pay for in cars comes down to two things: user experience and styling. That’s where most of the purchase decision happens. And increasingly, with millennials and other folks, it’s not even in horsepower or suspension, or all the things that automakers like to tout. It’s about whether you can connect to my iPhone, right?
It’s a very difficult environment, and very much a changing environment for automakers out there. And we believe that in order for them to meet the needs of the consumer, they need to focus their efforts, just like we’ve seen in cloud and mobile—companies focusing their efforts on the value that customers pay for.
D2P: What are some of the challenges that you’re encountering in trying to develop this type of platform?
JH: Some of the big challenges that we’ve encountered, certainly, are cultural. Automakers are at different stages of sophistication, across the industry, in terms of developing software for cars. Some of them are really advanced; some of them are really behind. And one of the big challenges is educating them and helping them understand the benefit of using software to implement functionality on their cars—and how to do that intelligently without wasting a lot of time.
Another big area is interfacing across all of these different types of hardware. I think that’s one of our big values. And the reason for that is that there are so many different types of hardware, and as you get further along the supply chain, it becomes more and more bespoke and unique and delicate. You end up working with different suppliers with one-off hardware that they built. There’s not good documentation, there’s no consolidated interfaces for their hardware, so you end up having to roll a lot of things from scratch. So it’s that fragmentation across the industry that’s one of the biggest challenges. It’s also where there’s an opportunity for us, but it’s still a challenge.
D2P: What would you say distinguishes PolySync from other platforms?
JH: It’s that decoupling of hardware from software, which is such a key thing to have. Suddenly, your software development program can run at software speed, and your hardware development program can run at hardware speed, which is slower, naturally, and it should be because the hardware is harder to make. Decoupling that, as well, means that you don’t actually need physical hardware to develop the software, which means that your contractors have it easier. You can work with startups to develop software, you can just buy software off the shelf, or you can reuse some from the public domain. There are so many benefits, and it’s that decoupling that is the key piece of value.
D2P: What has the customer feedback been like so far?
JH: Customers really like it. One of the big areas that we get used for, often, is simply that interfacing with different sensors. We make it so that you can plug and play all these different sensors that you see coming out of the market from all these different startups. You can plug and play those into our system, and your software basically doesn’t change at all. It’s completely decoupled.
D2P: Looking out into the future, how do you imagine the automotive value chain, or ecosystem, will look 10 or 20 years from now?
JH: If we’re successful, we’ll see automakers able to create a larger variety of different cars for different purposes. Some of them will be autonomous; some will not. But we believe that by utilizing technology from an ecosystem and increasing their sophistication in software development, they’ll be able to build cars in a fraction of the time and a fraction of the cost. I think that we’ll see cars get cheaper, I think we’ll see them used for different things, and I think that we’ll see a much greater variety across the industry.
D2P: What’s the coolest part of your job?
JH: For me, it’s been an interesting transition. I’m a first time CEO, and I’m an engineer by career and training, and I’ve come to lean a lot more toward the leadership side, being a CEO. The thing that really keeps me going is facilitating an environment where I get to bring in some of the smartest people that I would never get to work with otherwise, and create an environment where they get to do the things that they’ve always dreamed of doing. That is such a cool thing, and I hope I get to do it for a very long time.
D2P: What are some of the important things you look for when bringing a new person into your company?
JH: The first thing I look for, and the main thing, actually, is growth mindset. There’s a whole body of knowledge around why really high performers perform at the levels that they do. But basically, it comes down to a belief in human potential. A researcher at Stanford came up with the idea that people generally are either in sort of a fixed mindset, where they believe that human potential is fixed at birth, or a growth mindset, where they believe that human potential can be developed.
You may have heard people say “I just wasn’t born with a head for math,” or “I’m not artistic,” or “I can’t sing,” and similar things. That’s a very fixed mindset. And then you’ve also probably worked with people who say “That was pretty cool. I’d like to learn about that and figure it out.” And these people generally become better at things over time. These are people who thrive off of learning, and they love that, and growing and being better. And so that emotional connection to learning and growth is the first thing, and major thing, that I look for.
The second thing that we look for is, obviously, domain experience, specifically with what we need somebody to do. But, obviously, we’re in a new industry, and nobody has done this stuff before, and so a lot of the old patterns that we’ve seen in other industries just don’t apply. And so, the first thing I want is people who can learn and grow because there are challenges that we’re experiencing here that no one has ever experienced.