Canadian AI Opinion
Aug 09, 2018 ● Yemmit
Executive Interview: Dr. Foteini Agrafioti, Head Of Borealis AI And Chief Science Officer, RBC

Dr. Agrafioti is the Chief Science Officer at RBC and Head of Borealis AI

to Create Opportunity for Canada

Dr. Agrafioti is the Chief Science Officer at RBC and Head of Borealis AI. She is responsible for RBC’s intellectual property portfolio in the fields of artificial intelligence and machine learning. Prior to joining Borealis AI, Foteini was the Chief Innovation Officer at Architech, where she led Research and Innovation. She also founded and served as Chief Technology Officer at Nymi, a biometrics security company and maker of the Nymi wristband.

Foteini is the inventor of HeartID, the first biometric technology to authenticate users based on their unique cardiac rhythms. She is a TED speaker and serves on the editorial review boards of several scientific journals. Foteini was named “Inventor of the Year” in 2012 at the University of Toronto where she received a Doctorate in Electrical and Computer Engineering and was named one of Canada’s “Top 40 Under 40” for 2017. She recently spent a few minutes talking to AI Trends Editor John P. Desmond

Q. What is the background of Borealis AI?

A. Borealis AI is a new group within RBC that started two-and-a-half years ago. The objective was to create an institute where researchers can do fundamental machine learning research after graduating from academia. There are two reasons for this. One is that Canada has done well technically in the AI space, having developed deep learning and reinforcement learning foundational techniques, largely out of the University of Toronto, the University of Montreal and the University of Alberta. However, we have found that many of the graduates who had been on the forefront of publishing and creating the research were deciding to leave the country and pursue opportunities elsewhere. There were not enough interesting jobs in Canada.

We thought that AI could be a niche for Canada, so we wanted to create jobs that would be interesting enough to retain these people and to have them continue to do AI research in Canada.

We are a data-driven business, from our retail banks to the capital markets. We use data to drive decision-making; we use machine learning algorithms in our applications. So Borealis is the place where we build our intellectual property in this space.

We started off in Toronto and now have around 70 people with five research centers: Toronto, Montreal, Edmonton and in the fall, we are opening centers in Vancouver and Waterloo. Each is led by a professor who wants to continue to teach and to help commercialize the technology.

Each center will divide its time between fundamental and applied research. For fundamental research, we are looking to advance the state-of-the-art in machine learning and in supervised learning with a human-in-the-loop approach. These are hot academic research topics. This group is dedicated to pushing the boundaries of science and publishing the results of our efforts.

The applied effort is from a combination of researchers and software engineers. This is where we use the algorithms to develop software. We apply them to challenges in the financial services industry and beyond. We don’t have a sole mandate of delivering banking applications; we’re interested also in climate change and in healthcare, for instance. But the primary focus is on RBC applications.

Q. What is the role of the Edmonton center?

A. Edmonton is the first research center we opened outside of Toronto. The reason was twofold. First, the group there at the University of Alberta and the Alberta Machine Intelligence Institute had done extremely well in investing in reinforcement learning and in proving it can work. They have built a strong practice on that. We wanted to strengthen that community. They did not have many surrounding technology companies. We are a Canadian company and we wanted to see our community prosper from coast to coast, and we wanted to create jobs right in Edmonton to support the university.

Second, reinforcement learning is extremely interesting to us. It’s good for analyzing data. A lot of the data we deal with in banking requires pattern analysis. We hired Prof. Matt Taylor out of Washington State University to lead the Edmonton lab. We also work with private companies. We have an applied research practice in the financial sector.

The reason we expanded to Vancouver, which is a much larger city, was to create more opportunities for graduates. There is a strong computer vision specialty there, encompassing graphic and special effects, too. The University of British Columbia and Simon Fraser University have students advancing the state-of-the-art in video applications. We are expanding the data sources that we look to and are taking advantage of visual data.

Q. How many jobs do you anticipate being able to create in Edmonton?

A. Our goal is to create 160 jobs across the country by the end of 2019. We are at 70 today and we have 15 people in Edmonton at the moment. It’s a matter of finding the right people in each location. We don’t have a hiring number goal for each lab.

Q. How did you get into this work? What is your background?

A. I studied electrical engineering. It was at the University of Toronto that I got introduced to machine learning – biometric authentication in particular. We developed a way to authenticate people’s identities using the human heartbeat. Nymi was a company I built that commercialized that research. It was a wearable device. So, academia plus startups is my background.

I decided to join RBC because I had seen that path to commercialization and felt we could be doing it better in Canada. The US is very good at commercialization. I saw a lot of responsibility at RBC for nurturing the well-being of the country. Our interests aligned very well in that respect. I feel very passionate about it. It’s not straightforward to build groups like this within known technology companies. It’s very unusual.

“We do not yet have practitioners of AI.”

Q. What are the primary challenges?

A. I see two big ones. The first is recruiting talent, which is a very scarce resource, primarily because everyone who does AI today is coming from research, PhD programs and academia. We do not yet have practitioners of AI. That is really limiting when you are trying to build large teams, especially when you face the competition we do. There might be 20,000 people in the world who can do this job, and maybe 3,000 coming out of the universities each year. The competition is fierce.

The other challenge is that we are extremely regulated, for good reason. We have rules for how we can use AI to bring products to life. As you know, AI comes with concerns around issues including bias and privacy. These are the kinds of problems where there is active research and people are trying to solve these ethical problems. When we try to use these technologies in our products, we can hit walls because of how our industry functions. This is why we are so heavily invested in fundamental research. If bias in AI can be solved, we will be the first in line to make use of that.

We are enjoying the challenge. AI is moving so fast, and companies are adopting it so quickly. I don’t know if it’s good to always get into a race to bring new products to market. Sometimes it might make sense to step back and ask if the technology can be compromised. We have no other option to do business than the right way. So, it’s an engineering problem you have to solve. It constrains you, but it’s a good challenge.

Q. Do you have any advice for students interested in pursuing a career in AI for what they should study?

A. I would recommend computer science and machine learning. A career in AI can be very fulfilling. These technologies can be applied to many disciplines. Machine learning can be applied to finance, healthcare, consumer electronics, even the fashion industry. It’s very diverse. It’s exciting to be in this space right now. It’s challenging and you don’t have to limit yourself to one industry or field. This field will be pervasive as we move forward. We need more people who understand the core technology.

The other thing I would say is that students should consider coming to Canada to study. There is no better place in the world to learn. I have seen the support of the country for long-term investments in AI, such as in reinforcement learning, which has led to the creation of these new institutes [like Borealis AI]. We have attracted world-class talent from around the world. It’s a very good place.

Q. Where are you from originally?

A. I am Greek. I came to Canada for grad school. I found it cold, then I got used to it.

Q. Can you talk about one of your AI projects?

A. We have a project that uses AI to analyze news. Real-time information is important to our businesses. We make investment decisions based on it. If you ask a financial analyst what their job is, they will tell you, for instance, that their job is to know everything about healthcare in North America. But there’s too much information and every human is limited in what they can absorb, and what languages they speak. So you need to invest in technology to help.

Apollo is the product we built as a result of these challenges. It is an agent that reads the news for you and alerts you to information it thinks will be of interest. It can look at individual companies, see historical patterns and predict a likely change. It is using natural language processing, deep learning and graphic rendering.

To learn more, go to Borealis AI.

This article originally appeared in Techristic 

Article by: