Next Tesla in the B2B Market

Hanbin Lee transitioned from serving as a tank mechanic in the Korean Army to founding SEOUL ROBOTICS after realizing autonomous driving was fundamentally a software challenge. While passenger self-driving companies like Tesla and Waymo focus on public roads, Lee tested 100 industrial projects before narrowing his entire company down to solve the labor shortages hitting automotive factory yards. By mounting sensors onto facility infrastructure instead of individual vehicles, SEOUL ROBOTICS built the autonomous backend for BMW facilities and global logistics hubs. His journey highlights how turning away from crowded passenger markets to master one specific B2B bottleneck created a long-lasting industrial advantage.

Hello, my name is Hanbin Lee. I'm the captain of SEOUL ROBOTICS. 

SEOUL ROBOTICS is a B2B autonomous driving company. When you think about autonomous driving, there are companies like Tesla and Waymo that provide robotaxi services, autonomous driving you can use on public roads. B2B autonomous driving is not like that. We provide autonomous driving where there is a severe driver shortage.

We are currently focused on two areas. The first is finished-vehicle logistics. When a car is manufactured, somebody has to drive it to the ship, train, or truck. This requires 200 people per factory location for the carmakers, and there simply aren't enough drivers providing that labor. So we provide an autonomous driving system that moves these cars on their own, so they can be transported safely to your doorstep. The second use case is logistics centers. When you order a product, it goes through large logistics centers, and inside them there is a lot of manual driving. Somebody has to move the car to be charged, to be cleaned, to the second or third floor. We have 13 different nationalities on the team, and we've been focused on B2B autonomous driving for the last three or four years.

I went to the US when I was 12, to a small boarding school, and that's where I spent my youth. Then I went to Penn State. Until 12 I was in Korea, and my parents thought robotics was going to be the future, and this was the early 2000s. At Penn State I studied mechanical engineering, thinking that if I wanted to do robotics, I had to be able to build robots. But as I studied, I realized robotics is not a hardware problem. It is so much a software problem. So as soon as I graduated, I started studying software on my own.

Then I came back to Korea and went into the Korean Army, because I'm a Korean citizen. I spent two years as a tank mechanic, which was pretty fun. Right after finishing that duty, in 2016, I started looking into AI and self-driving with a group of people. I thought, the industry is new, they don't teach AI and self-driving in school yet, and this market is going to be massive. So why don't we start a company and see how it goes?

In August 2017, with four co-founders, I started SEOUL ROBOTICS, not knowing exactly what market we would go after. The company started from a study group on AI and self-driving. The study group entered a competition, and the competition team became the foundation of SEOUL ROBOTICS. I was the captain of that study group, and the title stuck. Still today, after eight years, my title is captain. That's the founding chapter of SEOUL ROBOTICS.

SEOUL ROBOTICS was my first gig, and it still is. We didn't know a whole lot. We knew our computer vision technology was pretty good, so our first approach was going after the OEMs, and for the first couple of years we really chased that. With the capital we had, we couldn't go after a full autonomous driving system and have a business model that generated revenue at the same time. It was going to be a very long-term, purely R&D effort. It turned out that carmakers move extremely slowly and are very conservative, and we realized that trying to sell our system to OEMs was just not going to work. So without them, what else could we do?

Since then, we did a lot of different autonomous driving projects, about 100 of them. Sometimes we added a bit more, control systems or camera systems, working as an outsourcing company, but a lot of these projects just weren't scaling. I got the feeling this was going to be tough to commercialize. There simply weren't enough forerunners who had shown that this kind of system reduces cost or improves efficiency by a significant amount, which is what it takes for B2B use cases to adopt the technology widely.

Our first major client was BMW. They had a problem so significant that all the way up to the C-level, they wanted it solved. When they make a car, somebody has to pick it up and drive it to the train station, anywhere from one to three kilometers, and it takes three or four shifts of workers driving these cars across a large factory. In automotive history, this has never been automated. It has always been driven by humans, and they simply couldn't find enough people to do it. This was one of our 100 projects. 99 failed, and this one succeeded. That's how we got our first glimpse of B2B autonomous driving.

If I had to explain B2B autonomous driving, Tesla is not going to provide autonomous driving for every industry segment. You use the Uber app or the Facebook app, but they run on AWS in the backend, and AWS makes a lot of money from all those companies. AWS is a B2B cloud system. Oracle or Salesforce maintain the databases for companies, and they're purely B2B backend companies. B2B autonomous driving is like that. We are domain-specific autonomous driving systems so that society can operate. We're the backbone, the backend autonomous driving system for society. That's how I like to think of what we do.

Our technology is quite unique and very different from Waymo and Tesla. First, we still predominantly use LiDAR. Second, the LiDARs are not on the cars themselves. They are installed around the infrastructure. You'll see poles with cameras and LiDARs around the facility. We have embedded our robotic system into the factory grounds, into the building itself. Along the path a car has to drive, we install these poles and sensors, and the cars are guided by the infrastructure. The software is identical to standard self-driving software that goes into a car. We just expanded on that idea to turn the infrastructure itself into a self-driving machine.

When you do that, you can solve critical issues that conventional methods cannot. First, you don't have to install hundreds of thousands of dollars of sensors per car. You need a handful of sensors around the infrastructure. Second, current self-driving can't handle heavy weather, because when it rains or snows the sensors are blocked by precipitation. With infrastructure, the sensors are installed much closer together, so you have a lot more of them, and you're brute-forcing the problem by putting more infrastructure into it. But it's not really about the hardware. It's about learning to fuse that immense amount of sensor data, processing it against the data we've collected over six years, and letting AI figure out what is rain, what is snow, what is noise, filtering all of it out to identify exactly where my fleet of cars is.

This technology, which we call Autonomy 3 infrastructure, solves one of the critical issues for B2B use cases almost as a byproduct. Cars need to move in fleets, groups of 5, 10, 50, or 100, and current autonomous driving technology wasn't going to solve that. But if the infrastructure orchestrates fleets of cars simultaneously, all of a sudden you can move fleets of cars safely. We didn't take the path of least resistance. We took the path that made sense to solve the problem step by step. And it turns out our system is extremely cheap to deploy. With the conventional method, you need to install systems on 100 cars. For us, it's a small parking lot, a handful of sensors, and you can move hundreds of cars simultaneously.

At CES, and especially at Nvidia, they're looking 10, 15, 20 years into the future and painting a picture of what it needs to be. I think they coined the term physical AI. The most breakthrough application is going to be robotics, and physical AI is simply AI agents going all the way to actuating robotic arms. What's the difference between a robotic arm and the engine of a car? It's the same thing. It's a self-driving system, or you could call it humanoid robotics. It is still a few years away from commercialization, but it's slowly getting there. What I admire about Tesla and Waymo is that they stuck to the problem, with the original design, longer than anybody else, with extremely talented groups of engineers.

Physical AI means more degrees of freedom in the robotics being rolled out for many different applications, with the humanoid robot as the epitome of that. Robotaxis and humanoids are similar in that they're B2C applications that can be widely used by everybody. But there are going to be a lot of sectors that need B2B autonomous driving products, and a lot of robotics for B2B use cases that won't look like humanoids, somewhere between the large industrial robots you see in car factories and the humanoid. A lot of B2B robotics is somewhere in between. It is going to be the future. It is fixed.

We pivoted several times before settling on B2B autonomous driving, and there was a lot of fighting inside the company. Out of 100 projects, which one do we keep alive? Some people wanted to keep several instead of dropping the other 99. I said no. We have to be the master of one single product and one single problem. We don't have the resources to dilute our focus. I think I made the right call. Those 99 products didn't matter. If you have one single killer item, it carries you all the way through.

I don't want to use Samsung as a bad example, but look at Samsung and TSMC, or Apple and Samsung. TSMC focuses on a single thing, Apple focuses on a handful of products, and Samsung does a lot of things. Look at how the market values those companies. TSMC and Apple have a higher valuation than Samsung. I think that makes sense. It was one risk with a massive upside. If I failed at this, it was okay, because the upside was massive, instead of a mediocre market where we spread our focus across different autonomous driving projects and were mediocre all around. That's the risk of starting a company. If we fail, we fail. But if you succeed in this one market segment, it is so massive that nothing could be better. As the company grew, people had different opinions about what products to build, multi-product strategies, multiple acts. I don't like that. That was one of the learnings.

Take Tesla. They started by focusing on the electric motor itself. They could build a powerful, efficient motor, so they built only that and outsourced the rest of the car, the battery, the body. Next was the battery, because battery technology was going to be very important for Tesla's future. Then they tried to sell a car using somebody else's body, and learned that you actually have to build the body frame around your core technology to build a solid product. Tesla outsourced some of its battery and powertrain technology in the early days, before the Model S, but that is building one success on top of another to verticalize the technology into a single solid product that outperforms everybody else.

SEOUL ROBOTICS took a similar approach. In autonomous driving, perception was the most important thing, so we went deep on the computer vision problem. On top of that we worked on many applications, but we found that to serve the B2B autonomous driving problem, we had to build the entire product, not just perception but planning, control, and the system around it. That verticalization was extremely important. And for it to work, you still have to own the core differentiation technology yourself. You can't outsource that. That's one of the problems a lot of existing carmakers have. Their core technology is not batteries, software, or electric motors. It's assembling parts. That difference in core technology is showing its impact today.

The autonomous driving we know is not limited to robotaxis. There's a lot to do in the world. The driver shortage is severe in Korea and Asia, and we need to provide robust autonomous driving technology to the backbone industry. It's still early days, and there's a lot to do. Right now we're one of the very few companies in the world providing autonomous driving technology to multiple carmakers and multiple logistics companies. I think we're going to be one of the long-lasting, very successful autonomous driving companies that solved this severe societal problem. We are based in Korea. If you want to solve autonomous driving problems across different industries, take a look at our website, SEOULROBOTICS.tech.