Alumni advice and artificial intelligence

Alumnus Kevin Swersky (‘07 BSc.) is no stranger to making good predictions. In fact, he’s built a successful career on machine learning, a subfield of artificial intelligence (AI) that focuses on models and algorithms that learn from data in order to make predictions. But when he looks back on his career, Swersky reflects that a little luck—and inspiration from the people around you—can go a long way.

Swersky is one of five team members that founded machine learning start-up Whetlab, focused on creating machine learning tools that help tune other machine learning systems, improving their predictive capability. Founded in 2014, the company was sold just a year later to Twitter. It’s a testament to the ever-growing importance of the field to industry, and the scale of the data challenges they face.

“Everyone recognizes the enormous potential that AI has in shaping the world, and there is a lot of focus on so-called ‘general AI’ able to tackle any task,” says Swersky, now working at Google Research in Toronto. “I think what people don’t typically realize is how small improvements in intelligence can open up incredible new innovations.”

Oftentimes, Swersky explains, there is a threshold of performance in which something goes suddenly from not working to working. For example, it was only recently that advances in speech recognition made the smart assistants used in phones and home devices possible.

“We’re seeing these kinds of thresholds being met in many other areas,” says Swersky. “Most exciting to me: AI is at the point where it is starting to have a real impact on other scientific fields, especially medicine. The same type of technology that we developed at Whetlab can also be used to help design new kinds of therapies. In essence, We’re starting to see AI and machine learning become a fundamental tool for science and engineering.”

From studies to start-up

“It was originally a desire to make video games that drew me to computing science,” says Swersky. “I knew that computing science would teach me the skills needed to achieve my goals, and during my degree I fell in love with machine learning—and that became the focus of my studies, and my career.”

Born and raised in Edmonton, Swersky says that he considered schools in other parts of Canada for a change in pace and perspective. 

“The University of Alberta has an excellent reputation for computing science though, so academically it wasn’t a difficult decision. The campus has a renowned artificial intelligence group, with a strong focus on machine learning. I had the privilege of taking classes from incredible researchers, whose own passion and excitement for machine learning inspired me to pursue it further. This shaped my eventual career.”

Founding Whetlab

The idea that would become Whetlab began not long after Swersky started his PhD studies at the University of Toronto, when a new brand of AI known as deep learning was beginning to emerge—a breakthrough supported by faster computers and increased ability to process large amounts of data. At the same time, Swersky explains, these systems were very difficult to get to work.

“This often came in the form of ‘knobs’ that controlled various settings of the system. Setting these knobs to the right values meant the difference between a system that worked and one that didn’t,” says Swersky.

Swersky and his colleagues found a way to tune these settings automatically with AI—using AI to tune AI—even beating out those found by experts on their own systems. And industry was starting to take notice.

“It was clear that we had developed a useful technology, and so we decided to form a startup in order to turn it into an automated service that anyone could easily use,” says Swersky. “The five of us met up in Boston one weekend in 2014, and Whetlab was born.

Just a year later, Whetlab was purchased by Twitter, joining Twitter’s core deep learning team, Cortex. There, Swersky and his colleagues integrated their technology and used it to help tune Twitter’s machine learning systems and improve performance of their computer systems.

“The acquisition itself was one of the craziest experiences of my life. For several months, I would have to fly to New York, Boston, or San Francisco on a moment’s notice and I couldn’t tell anybody why,” recalls Swersky. “Getting acquired is a rollercoaster ride, where one minute you think everything’s going smoothly, and the next the whole world is crashing down. Eventually, it’s done, the dust settles, and you move on to the next chapter of your life.”

Advice for the next generation

Swersky explains that his own career may have struck at the perfect time as the AI industry was just getting started, but the industry is far from slowing down—and he shares a few words of advice for students today.

“Take risks, but try to take calculated risks. For example, as an undergraduate, I took assignments extremely seriously and would always try to get 100 per cent. It meant that during the final exam, even if I was having a bad day, or made silly mistakes—which happens to everyone at some point—I could still do well in the course.”

For computing science majors, Swersky advises students to start thinking about their career early. For those wanting to pursue graduate studies, try to arrange research internships with professors early on. It's an invaluable experience that greatly helps with applications.

“If you persist against all odds, your odds of success improve dramatically. Surround yourself with the best and brightest people you can. I owe everything I am to the people who shaped me along the way. Make sure to find a group of people who challenge you and help raise you to their level. Do the same for them as well.”

“It was a privilege to have attended the University of Alberta, and I hope that it continues to be a warm and inspiring place for the next generation of students.”

Learn more about artificial intelligence research in the University of Alberta’s Faculty of Science.