Interview with Timmy Ghiurau, Software Engineer at Volvo — What it takes to become a self-driving car engineer

This is another great week for AI.

Elon Musk has recently announced Neuralink. He will put small devices into all our brains to help humans merge with software and move us a step closer to the Matrix.

Isn’t that great?

Today we are also at our third episode of the series of interviews with AI and ML experts. After Michael Green and Ola Lidmark Erikkson, we had the pleasure to talk with Timmy Ghiurau, software engineer at Volvo in the self-driving car department. He will talk about the skills required to become a self-driving car software engineer and how gaming is revolutionizing the self-driving car industry, starting from GTA 5.

If you have any questions, feedback, or you want to join our publication, just shoot an email at sp@journeyxp.com.

Can you tell me a bit about your story and how you landed a job at Volvo?

I moved to Denmark in 2010 to study in Copenhagen. It was the period when mobile apps started to bloom and that’s why I decided to join the new program Innovative Communication Technologies and Entrepreneurship at Aalborg university.

I studied different kind of technologies, software development practices, and finally I ended up in this incubator at Aalborg university during a startup bootcamp, where Eyetribe also had his office.

Volvo innovation center was also created in Copenhagen, a center of concept planning and innovation.

First, I worked at Eye Tribe and, then, I got approached by Volvo cars due to their interests in eye tracking technologies, computer vision and the skills that I have in AI and gaming solutions, especially Unity 3D.

I started my collaboration with them through my master thesis project and I started doing research inside the concept, planning and innovation center that Volvo has in Copenhagen. I was finally moved to Gothenburg where I got hired as a software engineer for the self-driving cars department!

What is the typical day in the life of a self-driving car software engineer?

It is always pretty exciting and fun to know that you are working with the latest hyped technology and helping to shape the future of the car industry in building more robust and safer cars.

Right now, my main focus is the infotainment system inside the car, but I still support the self-driving car team with 3D simulations and computer vision.

It takes a very broad knowledge of various technologies to work in this team: you work analyzing big chunks of data, observing patterns in simulations and different traffic situations, you need to have knowledge in computer vision, and also 3D simulation.

Every department inside the self-driving car division is really broad and demanding in terms of skills required.

You talked about how gaming was one of the reason that brought you to Volvo. How important is gaming right now inside the world of AI?

First, it strongly reduces the costs of real-world simulation, but it especially boosts the speed of the development of the self-driving technology.

Gaming is a great driver not only for self driving cars, but also for AI in general. It is a very good benchmarking tool where AI systems can perform tasks and get training. And games have a lot to do with learning and performing various tasks.

Testing and training AI in the gaming environment is a powerful tool but it is also dangerous, especially if you try to adapt the environment to the specific AI or you adapt the AI to perform at his best in one single environment. Especially in self driving car, AI needs to learn and adapt to any kind of situation to become better and better.

Again on this point, what do you think about Improbable and the work they are doing with 3D real-world simulation?

I think it is a really powerful tool and a huge step forward. Running those kind of scenarios in such a scale is really insane.

Using that kind of technology you can simulate the effects of accidents in traffic flows and using those data for anticipation. This would be a great enabler for self-driving car and would make them always a step ahead in predicting and avoiding accidents in real-life situation.

It would surely speed up the development and make it evolve faster

Self-driving cars are now at the peak of the hype. How do you think we should take the forecasts made by some analysts that see 2024 as the year of the first level-5 self-driving cars?

I think we still have a long road ahead. Car manufacturers have different opinions on this. Right now, level 5 is only theoretical. We will have to wait and see. I don’t really have a strong opinion on this.

Car manufacturers like Volvo, Ford, Audi are already adding autonomous systems technology to indoctrinate customers so that they can get used to it and adopt them gradually.

Currently, the main focus is making people familiar with these innovations. This is the reason why Volvo and Ford send around cars in London, Arizona, Pittsburgh etc. etc..

What do you see as the most critical hindrance for the development and adoption of self-driving cars?

It is predominantly a cultural matter.

People mentality is the biggest hindrance in this industry.

The way they open up to this technology is the key enabler. Today, if you check out the comments on video on Youtube or article on TechCrunch, you can see people that are looking forward to meeting a self-driving car in order to bully them and jump in front of them. This is why most of the car makers do not want special labels on the cars. If you put special tags on autonomous vehicles, they will be given different priorities and they will encounter different kind of behaviours, from which they will fast learn and react accordingly.

That’s why it is fundamental to indoctrinate people and teach them from early stage; starting from level-2, people can begin learning how they can rely on the car and gradually increase the feeling of safety.

You worked at Eye Tribe before joining Volvo and eye tracking and motion tracking was their core competences. How can this kind of technology improve the performance of a self-driving vehicle?

My main part of the research for Volvo was how eye tracking and motion tracking can enable self-driving cars. I was studying how much time the driver actually pays attention to driving and how much he spends doing secondary and tertiary tasks.

We are searching this area especially because the driver experience will be completely revolutionized.

The driver will become a passenger.

Hence, we are trying to build the car around the passenger to enable him to fully use his time.

Currently, the driver spends much of the time still driving, because doing other tasks more than 2.5 seconds can cause severe accidents. With ADAS (Advanced Driver Assistant Systems) and autopilot the time spent not driving can increase exponentially.

Eye tracking is a big player in this scenario. It supports drowsiness detection and always monitors if the driver is paying attention to the road. In this cases, the self-driving car can take control and make the driving experience much more safe thanks not only to external elements, but also internal sensors.

Transportation is an always-changing phenomenon. Do you see self-driving cars more suitable in an ownership-dominated world or more in a sharing-influenced scenario?

The self driving industry will become an essential element in the sharing-scenario. Nowadays, we lose a lot of time commuting and being stuck in the traffic. And very often most of the cars are filled only by the driver.

Transport-sharing self-driving car would make the traffic more fluid and the cars could be shared by more passenger sharing the same trip, having a huge impact also on the environment. In this way people could use their time much more effectively.

Automakers and ride sharing companies are focusing on this transformation. However, privately owned cars will still dominate the traffic for years especially in small cities and suburbs.

Most of the car makers are already in convergence with transportation providers and car-sharing companies.

What does it take to become a self-driving car engineer? Do you think it is something you can learn from a Udacity Nanodegree?

First, it takes a lot of passion and a broad knowledge on software development, AI, computer vision, machine learning, sensor programming, and also some hardware kwoledge.

A Udacity Nanodegree could provide people a strong skillset but I don’t think it is enough. You need to have some prior knowledge on software development, a strong mathematical background or important problem-solving capabilities to start.That made me succeed in a way.

Also having a gaming background can make developers better. They think out of the box, they are courageous enough to try new things, as gaming concepts that could help the AI systems to learn in a more efficient way.

This kind of Nanodegree can anyway enhance and structure your skillset and make you realize if you can work within this area.

Another element which is often disregarded especially by developers or engineer is the UX. Innovation makes sense only if it is made along the users. It is key to innovate in the right direction to find the real problem and give the proper solution.

What do you think about the skills that the Danish university system has provided you?

It provided me especially with strong skills in problem solving, researching, finding problems, and structuring my work. I also had great advice from some of my teachers.

I had mentors that really saw my potential and taught me not to constrain myself inside a single technology, thinking in a more interdisciplinary and innovative way.

What are your main sources of information and inspirations for AI and self-driving cars?

I usually attend conferences and webinars. I also read a huge amount of articles from Mashable or TechCrunch.

Three things you would like to tell to upcoming software developers or students that want to deep dive in the world of AI and self-driving cars?

I think they should be passionate about innovation and have a broad software development knowledge. Try to come with innovative algorithm all the time and not be afraid to think out of the box. There is even someone which right now is using GTA 5 to train AI for self-driving cars!

Do not be afraid to combine technologies and never limit yourself to one single thing.