An interest in coding and a professor passionate about his work ignited Amélie Rolland’s interest in computer science, leading her to find ways to use AI for the greater good. From working on therapeutic drug discovery in a research lab to treating contaminated soil and water, Amélie is focused on using the latest AI technology to solve global problems.
What got you started in technology?
I played a lot of video games as a kid. I liked to do speed runs of games like Super Mario World and Conker's Bad Fur Day, but I was also a big fan of story-rich role-playing games (RPG) like Final Fantasy IX. When I was around 17, I wanted to build my own game. So I started to use an RPG game development engine. The engine didn't require any coding skills, but it was much more powerful if you knew how to code. At some point, I wanted to customize things, so I ended up copying some code online and editing it slightly to do what I wanted. I never finished the game, but I had a lot of fun working on it. This project was just enough to give me an idea of what programming was, and I decided to pursue computer science after that.
Tell us about your background.
I started in the field by earning a bachelor’s degree in computer science. Before I got in the program, I talked with computer science students and professors from different universities in Quebec. But when I went to the open day at Laval University in Quebec City, that's when it really clicked. I met Mario Marchand, a professor and machine learning researcher who was the program director at the time. Mario is the kind of scientist so passionate about what he does that it kind of becomes contagious. He started to talk about his research on drug discovery and how machine learning could help in the development of new medications for different diseases. I think at that point he almost convinced my dad to join the program too. I ended up choosing to study at Laval University.
My favorite courses in the program always involved some kind of practical coding project. I really enjoyed my internships in Web development as well, but I still thought a career in AI would be even better. During my last year in the program, I started to talk about AI with one of my friends who was doing a master's degree in machine learning. He later introduced me to his research lab, the GRAAL, and that's when I met Mario a second time. The GRAAL is a research lab that focuses mainly on learning theory, computational biology, and natural language processing. I started by doing an internship in the lab under Mario's supervision. I really liked my experience and decided to do a master’s degree co-advised by François Laviolette, another professor of the GRAAL who is now the director of the Big Data Research Center at the university.
Tell us about your experience working in a research lab.
Being part of a research lab was an incredible experience that helped me develop skills in machine learning and scientific research. I got the opportunity to work on my own research project and also on other machine learning projects that were done in collaboration with companies like Thales and Desjardins. I think it's really important to choose your research advisors well, but the other students in the lab are as important because they're the ones you'll be spending most of your time with. I ended up spending a lot of time working with Sébastien Giguère. Sébastien was doing his Ph.D. at the time, but he's currently the co-founder and head of research at InVivo AI, a start-up that helps to develop new drugs.
We worked together on the problem of machine learning-based drug discovery. In drug discovery, the first step is to identify a target protein that is responsible for a specific disease. After that, the goal is to find a molecule that can bind to the target protein to produce a therapeutic effect on the disease. In the machine learning setting, we had access to a training set of peptides (small proteins) each having a real value (binding affinity) that represents how much they can bind to the target protein. We first trained a machine learning model on this set that can receive as input a new peptide and output a predicted binding affinity. We could then send a list of peptides to this model to determine which ones had the highest predicted binding affinities and were the best candidates to be synthesized in a laboratory. The problem is that there is an exponential number of possible peptides, so predicting all their binding affinities to find the best ones would take years. For that reason, Sebastien and I worked on developing a branch-and-bound search strategy to find the most promising ones. The approach works by using bounds on sub-spaces of peptides to orient the search toward the peptides expected to have the highest predicted binding affinities. At the 2015 International Conference on Machine Learning (ICML), we co-authored a paper on the subject that was published.
After my master’s degree, I joined a start-up that uses AI to help solve environmental challenges, WikiNet.
As a woman who works in technology, do you have advice for other women considering this field?
I actually think that my life has been easier as a woman in a field dominated by men. The main reason is that being a minority makes it easier for people to remember you, and I think standing out is important for your career development.
At some point during my bachelor’s degree, I met a student one year ahead of me whom I had never seen before. I introduced myself, but he told me he knew who I was because I was one of the two girls to join the program the year before. When I go to computer science conferences, I always remember the women I meet more than the men, because there are a lot fewer of them.
My point is that, as a woman, you're going to stand out whether you want to or not, and I've always seen this as an advantage. If you have the skills, people will think of you when they have open positions to fill or when they need someone to speak at their conference. Of course, my own experience is not representative of every woman's experience, but I think the media are focusing a lot on bad experiences, while mine has always been positive.
What projects are you working on now?
I'm currently working on a project at WikiNet that uses AI to support the treatment of toxic substances in soil and groundwater. Basically, a lot of human activities like mining, agriculture, and other industrial activities can lead to the contamination of soil and water. An example is the TransCanada Keystone Pipeline that leaked over 700,000 liters of oil in South Dakota last year. This contamination can have negative impacts on the environment, but also on our health since the high exposure can increase the risk of cancer and other diseases. The problem is also more widespread than we may think. There are over 200 million people that are potentially affected by toxic substances from contaminated sites in 50 developing countries, according to Pure Earth, a non-profit that helps clean up contaminated sites in the poorest communities. Many developing countries have also less experience in cleaning up contaminated sites because they don't see this problem as a priority, or haven't until very recently.
So our goal with this project is to learn from past cleanups to provide recommendations for treating contaminated sites worldwide. We use AI to extract relevant information from environmental site reports and recommend treatments for contaminated sites.
Both problems are very important as both past cleanup results and treatment predictions can help to guide future decisions and transfer knowledge among experts. This can result in cost improvement and timeframe reduction, which are both needed to improve the number of contaminated sites cleaned up each year. With this project, we are competing in the $5 million IBM Watson* AI XPRIZE* competition, a four-year international competition that aims to demonstrate how humans can collaborate with AI technologies to tackle some of the world’s greatest challenges. The competition was launched in June 2016 and will end with three finalists participating in the grand prize competition on the TED2020 stage. WikiNet has been selected as one of the top 30 teams that are advancing to the third round of the competition.
Tell us about a technology challenge you’ve had to overcome in a project.
One of the hardest challenges we faced for our project at WikiNet was the lack of labeled datasets in our field. Pretty much the only data we had access to were environmental reports, but our machine learning approaches needed datasets to be labeled so that we could know, for example, which treatments were applied for a specific contaminated site. We tried several different strategies, always trying to improve our current labeling process. We now have our own team of environmental experts that labels our datasets and ensures that our machine learning models are trained on clean data.
How are you planning to leverage Intel® technologies in your work?
I'm particularly interested in the Intel® AI DevCloud to run machine learning experiments on Intel® Xeon® Scalable processors. At WikiNet we use different machine learning algorithms that must constantly be trained and evaluated on new data. We also have CPU-intensive tasks like converting large datasets of PDF reports into text data. I think we could leverage the compute power of Intel AI DevCloud to run more experiments, get faster results, and speed up our dataset preprocessing steps.
At the 2018 Neural Information Processing Systems (NeurIPS) Conference, I had the chance to meet machine learning researchers working at Intel. Nesime Tatbul, Justin Gottschlich, and their team at Intel propose a new way to evaluate the accuracy of time series classification algorithms that are motivated by anomaly detection tasks. I think many anomaly detection problems in environmental research can occur over time, which makes the new metrics extremely useful to us. For more information, see Precision and Recall for Time Series.
What trends do you see happening in technology in the near future?
I think the adoption of AI in enterprises will keep increasing over the next few years. I can see it happening in Montréal and Quebec City, with many new open positions for AI roles in companies that were not doing AI before. At the same time, there seems to be a lack of qualified resources in the field. So this might slow down the whole process, but new training programs in AI should help to increase the number of qualified candidates. There also seems to be more and more consulting companies focusing on AI. So if enterprises are not able to fill their AI positions, they might still hire consultants.
Outside of technology, what type of hobbies do you enjoy?
I still like to play video games. The last RPG game I finished was Final Fantasy XV, and I also play Rocket League from time to time. I recently bought an Oculus Rift* and I can't stop playing Beat Saber. I also like board games. I play Dungeons and Dragons (D&D) with my friends and just started to play Betrayal at House on the Hill Legacy. Other than that, I go rock climbing every week and I like to read both fiction and non-fiction books.
Learn More about the Intel® Software Innovator program
Check out the latest innovator updates, get the program overview and learn about Innovator benefits. For inspiration or to start sharing your work, join our community collaboration platform Developer Mesh.
Interested in more information? Contact Program Manager Wendy Boswell.
Intel's compilers may or may not optimize to the same degree for non-Intel microprocessors for optimizations that are not unique to Intel microprocessors. These optimizations include SSE2, SSE3, and SSSE3 instruction sets and other optimizations. Intel does not guarantee the availability, functionality, or effectiveness of any optimization on microprocessors not manufactured by Intel. Microprocessor-dependent optimizations in this product are intended for use with Intel microprocessors. Certain optimizations not specific to Intel microarchitecture are reserved for Intel microprocessors. Please refer to the applicable product User and Reference Guides for more information regarding the specific instruction sets covered by this notice.
Notice revision #20110804