Automakers dive into new talent pools

Industry scrambles to find engineers for self-driving programs

March 06, 2017 12:00 AM

Fifteen-year old Mikel Bober-Irizar opts to study autonomous driving over hanging out with friends.

At age 15, British computer whiz Mikel Bober-Irizar cuts an odd profile as a student leader in an online self-driving car engineering course.

In the semianonymous world of e-learning, he codes alongside Ph.D.s and veteran developers from around the world for hours on end -- even if it means frustrating his friends by choosing to do lessons over hanging out with them.

"My dad is very much into computer science. He got me my first laptop when I was 5 years old," said Bober-Irizar. "I've learned to teach myself things, figure out how to do them by just doing it."

Not bad for a kid who can't even drive.

Bober-Irizar's interest in autonomous vehicle technology doesn't come from a childhood of playing with Hot Wheels, but from a deep interest in computer science honed through competitive online challenges.

He's one of countless engineers and hobbyists who lack what one recruiter calls "pedigree experience" but offer a potential source of new talent for companies desperate to staff rapidly expanding au-tonomous vehicle programs. Today, automakers and suppliers are stymied because the academic institutions and programs needed to train autonomous vehicle developers are either slow to adapt or nonexistent.

"We don't have enough programs, we don't have enough courses and instructors, and we're just one industry," said Kristin Dziczek, director of the Industry, Labor and Economics group at the Center for Automotive Research in Ann Arbor, Mich.

The course Bober-Irizar is enrolled in, run by e-learning company Udacity, has received 20,000 applications since it launched late last year. It has accepted a quarter of those applicants. While many of the students may lack the resume one would expect to work on self-driving vehicles, they possess a fascination with artificial intelligence and a belief in using technology to improve the human condition.

"It's not about the self-driving car," said Bober-Irizar. "It's about artificial intelligence, automation. I want to do something that helps the world."

■ Rapid growth

Companies must seek out talent, not wait for applications to roll in

Sebastian Thrun, founder of e-learning company Udacity, says companies looking to hire autonomous vehicle engineers are hampered by the fact that only a tiny segment of the work force has training in the area.

Aggressive targets to bring self-driving vehicles to market have left the industry scrambling for engineers with skills in robotics, artificial intelligence, machine learning and related fields.

"The biggest challenge is the speed and the rate of growth," said Sherif Marakby, vice president of global vehicle programs at Uber who spent 25 years at Ford Motor Co. "The size of the pool is very good, but at the same time, the growth rate that we need still exceeds the capability to recruit people."

Companies have responded by cultivating Silicon Valley-like cultures, betting on startups, such as Ford's billion-dollar investment in Argo AI, and spending lavishly to court talent. Depending on skill level, engineers in this realm can earn $100,000 to $200,000 a year. Competing on compensation, though, only goes so far.

"Financial capital isn't the constraint. It's the talent. It's the knowledge worker," said Bill Huffaker, global director of talent acquisition at General Motors.

Huffaker compares headhunting for hard-to-fill fields such as artificial intelligence to how "a supply chain would source materials."

Last year, GM brought its recruitment largely in-house. It started by having engineering leads assess possible candidates, and built an internal database by scouring sources such as LinkedIn and conference speaker listings for potential hires.

With such high demand, recruiters can't wait for applications to roll in, Huffaker says.

"It's very different these days," he said. "Roughly 25 percent of people are looking for jobs. Seventy-five percent of candidates, you need to go after them."

The main source for talent starts in the university system.

"One of the biggest challenges for companies looking to hire self-driving car engineers is that there is only a tiny segment of the work force that's already been trained in this area," wrote Sebastian Thrun, founder of Udacity and Google's original autonomous car program, in an email interview.

■ Ground zero

Carnegie Mellon tries to deal with a 'flood of companies'

Carnegie Mellon career fairs attract recruiters from many companies, including Uber, which raided the university's robotics faculty to launch its autonomous vehicle program.

A prime spot to witness the talent war firsthand is at university career fairs, where physical constraints can often pose the biggest challenges.

"There's only so much room that we have," said Kevin Collins, a career adviser in Carnegie Mellon University's computer science department who works with companies looking to recruit its students. "A couple years ago, we doubled our interview rooms," said Collins. But since then, "there's been a flood of companies interested in talking with our students. We're booked."

Companies such as GM and Delphi have shown up to scout talent from one of the best robotics programs in the world. Uber famously raided Carnegie Mellon's robotics faculty to launch its autonomous vehicle program.

The field may account for only a slice of the computer science majors at Carnegie Mellon, but even in this small area, competition is fierce. The same skill sets required of an engineer working on self-driving vehicles could be used at Amazon to improve its complex logistical operation or at NASA's Jet Propulsion Laboratory to put rockets in space.

Of the 283 bachelor's, master's and doctoral students who graduated from Carnegie Mellon's computer science department in 2016, just six went to work in the auto industry, according to a self-reported survey of graduates provided by the university.

"The challenge for [automakers] is to compete with the Microsofts and the Googles. It's a smaller pool of people and ever-increasing number of companies," said Collins.

Collins: “There's only so much room."

But a rapid expansion of university-level programs may not happen soon.

Robotics is still a nascent field, meaning top-quality programs need to develop coursework, find instructors and attract students.

"A new program is going to start out as a cross-disciplinary degree," said CAR's Dziczek. "It's evolutionary, and it's also marketing to the student body."

For example, Carnegie Mellon's Robotics Institute launched a master's program in computer vision that just graduated its first class of 24 students at the end of 2016, out of an applicant pool of nearly 500.

Martial Hebert, director of the Robotics Institute, said he believes that number can grow significantly.

"I think that's true across the board," Hebert said of the field's evolution. "On the undergrad side, we're not anywhere close to what we need."

Hebert sees a corollary to other groundbreaking automotive technologies.

"HVAC used to be reserved to very, very highly skilled individuals," he said. "Then, as the field matured, you need more technicians."

"In that sense, we know where it goes. We need to produce a lot more."

■ Looking within

Automakers weigh time necessary to retrain their own engineers

Marakby: Uber needed tech integrators.

Where will self-driving vehicle engineers come from?

In some cases, companies are finding talent within the industry.

"A lot of the engineers that we have did not start in robotics," said Uber's Marakby.

Engineers who were trained to work in manufacturing, product development or other automotive concentrations are finding opportunities in autonomous vehicle development. A background in cars might even be an asset.

"At Uber, there was already a strong software team and a strong hardware team," Marakby said. Instead, they needed engineers "that understand the auto industry, that understand how to integrate technology into the car."

But helping engineers make that transition can be time-intensive.

"There are plenty of really talented engineers [at GM] that want to make a career transition," said GM's Huffaker. "Do you go externally? Or can you afford to slow down and train someone?"

Ford has partnered with a software consulting firm to ease such transitions. Pivotal Labs, headquartered in San Francisco, creates programming for industry engineers to transition to new fields and uses the best practices from startup culture.

■ Applied tech or theory?

An ophthalmologist designs lenses for General Motors

Udacity's courses are designed to give students experience as well as theoretical knowledge.

Automakers can also work with universities to encourage the type of engineer they desire. Audi partners with 30 universities worldwide on research and directly sponsors doctoral projects to expand the pool.

Apprenticeship-style programs similar to what may be found in Europe are tantalizing for companies looking to connect with students during their education.

"In general, applied technology is the key," said Michael Hafner, Mercedes-Benz's director of driver-assistance systems. "We see people coming from universities with more experience than just theoretical knowledge."

Udacity's online self-driving engineer coursework is built on this premise. Students run through lessons targeted at specific autonomous functions, such as lane-keeping, and learn how to program to address those challenges.

"It's not ridiculously difficult, but it does take time and effort," said Jessica Yung, a University of Cambridge economics student enrolled in the online course. "I'm more aware of the specific challenges involved with self-driving cars now that I know the mechanisms through which people are programming."

To what degree universities embrace this type of learning on a wider scale is an open question. Carnegie Mellon's Hebert cautions against focusing too much on practical coursework at the expense of foundational theory.

"We know what is used now. We don't necessarily know how that's going to evolve," he said.

"The important thing is for people to not only learn the tools, but to learn how someone arrives at those tools. Then you can adapt to changes."

He points to the increased use of a technique called "deep learning," through which large data sets of images are acquired through test drives to improve a system's autonomy, as an example of how the field can evolve.

Rather than undercut theory-based education, Hebert hopes to draw potential robotics students from wider academia -- from mechanical engineering and mathematics to biology and neuroscience and even social sciences. One recently created course recruits neuroscience and robotics students to research human-machine interactions.

"Robotics is one of the most multidisciplinary fields," said Hebert. "The spectrum is huge. That makes it exciting, but it makes it difficult to convey."

Huffaker: “Army of experts"

Focusing on the variety of disciplines needed can be a strategy for recruiters as well, seeking individuals who can specialize rather than those who have full mastery of several areas.

"It takes a small army of experts," said GM's Huffaker. One addition to the carmaker's self-driving team is an Israeli ophthalmologist who wanted to make a transition and is now working on lenses for autonomous technology.

■ DIY qualifications

Automakers look beyond coders with 'pedigree experience'

The apparent disconnect between the industry and academia, coupled with the rise of open-source, collaborative development online, is an opportunity for companies to experiment with the standard recruitment pipeline.

"We do see a growing number of folks who are very active and have very sophisticated skill sets in coding, software engineering, robotics, but don't necessarily have the pedigree experience," said Kiersten Robinson, Ford's executive director of human resources for the Americas. The carmaker has been reaching out to these people through informal meetups and events where coders gather.

Carmakers have known of the potential of open-source development for some years now, though mostly in the realm of app creation and infotainment. The spread of software to core vehicle functions, however, has created opportunities that resemble the world of gearhead culture with a slight twist.

"What we've seen is that the community is profoundly powerful," said Josh Hartung, CEO of PolySync, a company that provides open-source software to develop self-driving vehicle systems. "You've got tuners now that are remapping electronic control units and making them faster just with software."

Inherent to this do-it-yourself mentality, however, is a belief in the common good of shared technology that cuts against the grain of an industry locked in fierce competition.

One can glean this from an online group chat organized by Udacity that has become an informal hub to exchange news, tips and thoughts on the autonomous vehicle industry.

In December, a discussion erupted around George Hotz, the inventor behind an aftermarket kit that adds semiautonomous capabilities to select vehicles. After sales of the kit had to be stopped because of safety concerns expressed by the National Highway Traffic Safety Administration, advocates and critics of open-source tech spoke up.

One user, critical of Hotz keeping some technical knowledge private, wrote, "I prefer to spend my time on something that many could use and advance faster the self-driving car space, and the more open it is the better."

When another opined that the success of this technology remains in private industry, namely with Tesla, the response was as enlightening as it was brief.

"I hope not :-)," the responder typed. "The success of humanity should be the result of everyone, not one company!"