L-r, Alley Heaps Undergraduate Research Internship award recipients for 2025 are Devin Smith, John Kendall, Kate Redfern, and Elizabeth MacDonald.
Receiving the Alley Heaps Undergraduate Research Internship has meant so much to four StFX computer science students. The students say their summer in research has been rewarding, has deepened their knowledge, and has been pivotal to providing a clearer career path.
Devin Smith, John Kendall, Kate Redfern, and Elizabeth MacDonald each received $9,000 for summer research work under the supervision of a StFX faculty member. The internships are part of the Dr. H. Stanley & Doreen Alley Heaps Chair, which provides for the support, exploration, and advancement of computing science at StFX.
The students spent their summers working on projects ranging from developing software with medical and business applications to training a virtual self-driving race care to navigate safely around a track using reinforcement learning.
CANNOT EMPHASIZE ENOUGH THE VALUE
"I cannot emphasize enough how great of an experience I had this summer. I consider myself fortunate to have been selected as one of four recipients for the Alley Heaps Undergraduate Research Internship," says Elizabeth MacDonald of Antigonish NS, who is entering her final year of study in the Post-Baccalaureate Diploma in Artificial Intelligence.
She is supervised by Prof. Jean-Alexis Delamer.
"This experience has deepened my interest in artificial intelligence. The technical and problem-solving skills I gained will be directly valuable in future research projects and in my career," says Ms. MacDonald, who over the summer worked on training a self-driving race car, inside a virtual environment, how to navigate safely around a track using reinforcement learning.
The simulation software she used is called F1TENTH (F110), which models a small Formula-1 style car at 1/10 scale.
"The agent acts as the vehicle's driver, and the agent learns from the outcomes of actions. Agent choices are guided by a policy, which is like a decision-making strategy that improves as the agent continues training," she explains. "The agent tries both good and bad driving actions and learns which ones work better over time. A structured system of rewards and penalties tells the agent what a desirable or undesirable action is. For example, driving forward earns rewards, while reversing or colliding results in penalties. The overall goal is to train the agent to navigate the track and complete laps efficiently and without collisions. By the end of summer, Prof. Delamer and I were able to get the agent to navigate laps during training."
Ms. MacDonald says checkpoints were added to the base F110 environment to mitigate undesired agent actions. "I was able to implement an autopilot system to confirm that lap and checkpoint logic were working as expected. Prof. Delamer and I had an agent that was making improved decisions with these changes to the base environment. We are training the agent on a single track for now with the goal of incorporating different racetracks into the training regimen. Depending on the outcome of training, we will be evaluating how well the agent navigates tracks that it has not previously seen."
She says it was rewarding to watch the simulated car successfully complete a lap. "Aspects of the autopilot feature that Prof. Delamer and I developed were used to shape the agent's reward structure. Overall, I had an excellent experience working in the field of reinforcement learning, and this summer solidified my interest in artificial intelligence.
"This opportunity allowed me to apply concepts I learned in the classroom and to expand my knowledge by learning and applying new concepts throughout the summer. I am grateful that I had the chance to contribute to a project in the fields of artificial intelligence, reinforcement learning, and autonomous racing."
PUBLISH PEER-REVIEWED RESEARCH
"I am grateful for this opportunity and the support that allowed me to spend the summer doing research in machine learning. It was rewarding to apply my education to produce software with medical and business applications and publish peer-reviewed research as a first author," says John Kendall of Antigonish, NS.
Mr. Kendall is a fourth year student, supervised by Dr. Jacob Levman, completing an advanced major in computer science and mathematics.
This summer, he worked on making advanced machine learning research more practical and usable outside of the lab.
"Dr. Levman and his students use a program called df-analyze to test many different machine learning methods and figure out which ones work best for different problems. My role was to create a companion tool, called df-deploy, that takes the results from df-analyze and turns them into models that can actually be applied in real-world settings. I also integrated the software of one of Dr. Levman's graduate students into df-deploy so the models can handle image and text data. Df-deploy can serve as a tool with applications in many fields such as aiding medical professionals in diagnosis efficiency."
Mr. Kendall says they finalized and published a paper titled "Machine Learning and Feature Selection in Pediatric Appendicitis" in the peer-reviewed open access journal, Tomography. His software, df-deploy, is being prepared for use in research projects, including those in Dr. Levman's courses, where it will help students and researchers apply machine learning models in practice.
"This project has strengthened my CV with both software contributions and a published paper, which are major assets in applications for graduate school and future jobs. This experience has led me to consider pursuing further studies in graduate school."
WILL HAVE MAJOR IMPACT
Devin Smith of Antigonish, NS, who is going into his fourth year of a computer science degree, and is supervised by Dr. Taylor Smith, focused his research on ray tracing, a branch of computer graphics aimed at rendering images.
Ray tracing simulates real life light rays inside a digital scene in order to render a proper image. During his research, he used software such as Unity to build a ray tracer from the ground up, hoping to optimize its performance using different data structures, and various techniques.
"Ray tracers are notoriously slow, requiring a lot of computational resources to produce a realistic image. The main research topic in mind was to experiment with the ray tracer and to find ways to make ray tracing faster and more performant."
Mr. Smith says the experience felt a lot more independent and advanced than previous jobs he's held. "This opportunity means quite a lot to me. Personally, I plan to pursue computer graphics as a career path in the future. Unfortunately, StFX does not currently offer any courses related to computer graphics. This means that, up until this point, I have been entirely learning about this branch of computer science through online resources and my own research. To be able to fully dedicate my work to researching and learning about computer graphics is a huge opportunity, both to further my knowledge on the subject and to bolster my chances for a future career, and I am immensely grateful for that," he says.
Since he plans on finding employment within this branch of computer science, he says the experience will have a major impact on his future studies and career. "This research acts as both a demonstration in my resume that shows I have experience, and also a way for me to learn more in depth about the subject. I also think that it was able to show me a glimpse of how working within computer graphics would feel as real employment. Again, I am very glad to have been able to perform my research given how pertinent it is to my aspirations."
OPPORTUNITY TO EXPLORE RESEARCH
Kate Redfern, a third year student from Sherwood Park, AB, who is supervised by Dr. James Hughes, says this opportunity meant so much to her.
Her research had a few twists and turns over the summer. It eventually landed on a project using evolutionary computation, a form of artificial intelligence inspired by the natural process of evolution, to improve results from reinforcement learning algorithms, another form of artificial intelligence.
"Mostly it was an opportunity to explore whether research is something I am truly interested in. School assignments are very structured in the methods, tools, and expected results, but in research there are a lot more unknowns to start, and there is a lot more detail work. You're not always working with someone else's data or code, sometimes you're starting from scratch, building your own. You even have to come up with the questions to ask in the first place, which I really enjoyed," she says.
"This project has influenced the direction I take my honours thesis. I've already made some unexpected detours from where I thought my degree was headed, so this isn't a big surprise, but I wasn't expecting research or a graduate degree as my next step. Yet those are definitely on the table now!"
The end goal of this project is not just to improve the results of reinforcement learning, she says, but also to increase efficiency by decreasing the run times and overall energy usage. She hopes to continue working on it over the school year.
Further work will look at other applications for this method, beyond the limited scope of the environments they've looked at so far, as well as refactoring the code.