���ϳԹ���Դ

155 ���ϳԹ���Դ students were awarded the Undergraduate Student Summer Research Fellowships (USSRF) and Undergraduate Student Research Assistantship (USRA) awards this summer.

Students worked full-time for 16 weeks on their research projects, publishing their abstracts in the Inquiry@���ϳԹ���Դ Journal this month, an undergraduate research publication.

While students receive funding through ���ϳԹ���Դ, their research often takes them outside of Kingston. USSRF recipient Lucian Mussells, ArtSci ’27, travelled to Bruges, Belgium, to research Jan Provoost, an early Netherlandish painter who worked in the 15th and 16th century, for a planned exhibition at the  in 2029.

Other students travelled closer to home — studying  at Hopewell Rocks Provincial Park in New Brunswick or to ���ϳԹ���Դ Biological Station (QUBS) for fieldwork.

To Bruges and back

“Bruges was a center for a long time for so many kinds of art. Some of the most famous artists in the early Netherlandish painting world worked in Bruges,” said Mussells, a fifth-year Art History major, in an interview with The Journal. Mussells spent the summer working on the first monographic exhibition on Jan Provoost.

He became engrossed in Early Netherlandish Art while taking a course taught by Professor Ron Spronk, a leading expert in technical art history, two years ago.

“I’ve always loved it,” Mussells said, “but it was really Professor Spronk’s classes that sparked the passion that I have for it today.”

Mussells began the summer in Kingston, reading and digitizing literature about Provoost collected over several decades — the rest of which was in Europe. He flew to Ghent in June, where he stayed for the month, splitting his time between the Ghent University Library special collections and the BRON Research Centre in Bruges, in the Musea Brugge.

“A lot of that was a continuation of what I had started in Kingston, seeking out all of this literature and cataloging it, scanning it, reading it, seeing what its relevance is to the exhibition,” Mussells said. Only a fifth were in English, so his knowledge of Provoost’s work and Netherlandish art helped him recognize what to look for.

“He is a very interesting case,” Mussells said about Provoost.

Historians think Provoost began as an apprentice in the workshop of a famous painter, Simon Marmion, and when he died, married his widow to keep the business going, as was customary in workshops. This allowed him to expand his workshops to Antwerp and Bruges, and travel widely — including to Jerusalem.

“He becomes a member of the Brotherhood of Jerusalem Pilgrims in Bruges, which is this prestigious confraternity in Bruges, located in the Jerusalem Chapel, and it sets him apart from many other painters and artists of his time,” Mussells said.

Provoost is listed as a painter and a knight in the archives, likely having been knighted in the Order of The Holy Sepulchre in Jerusalem after the pilgrimage.

His paintings are scattered across Europe, and studying how they got there is an ongoing subject of research, Mussells explained. “That’s part of the research I’ve been doing, looking at networks of pilgrims and pilgrimage routes.”

Provoost also did not sign any of his paintings, leaving historians to attribute his work through a process of elimination.

“His body of work for a long time was sort of a wastebasket that paintings that didn’t have a solid attribution could sort of be thrown into” called the Provoost group, Mussells explained.

It is difficult even using technical tools to attribute works with early Netherlandish painting as well, since there is not one painter but many. Another goal of the exhibition will be to reappraise works attributed to Provoost and clarify the Provoost group.

“The exhibition also coincides with the anniversary of the Jerusalem Chapel that the Brotherhood of Jerusalem Pilgrims had their operations in,” Mussells said.

The Chapel was the site of a huge altarpiece by Provoost, whose panels are housed in different institutions, which Mussells hopes will become part of the exhibit.

“People don’t realize that paintings and our works of art are these physical, three-dimensional things. And by traveling and having these experiences, you get to see these things as the functional objects they are,” he said.

Mussells looks forward to continuing his work with the exhibition as he plans to pursue a masters in Europe.

Tiny mussels, big problems

Khanyiso Mnkantjo, ArtSci ’28, a third-year civil engineering student, spent the summer experimenting with nanobubbles — extremely tiny gas bubbles that remain stable in water for weeks — to deliver oxygen to a bacterium that is lethal to invasive zebra mussels in the U.S. For this work, Mnkantjo received an NSERC USRA.

Zebra mussels are wreaking havoc on freshwater ecosystems, Mnkantjo explained in an interview with The Journal. They are “filter feeders,” which means they take in large amounts of water to consume the food particles within it and expel the rest, clearing the surface of the water.

This leaves less food for other species and allows sunlight to penetrate deeper, disrupting the aquatic food system and warming the lake. This pushes cold‑water fish deeper, altering migration and spawning patterns.

Despite being only one to four centimetres wide, each mussel can filter up to a litre of water a day and they reproduce rapidly. “They just decimate our waterways,” she said.

Zebra mussels also cling to hard surfaces, affecting infrastructure like pipes and dams, as well as the larger native Canadian mussels, which suffocates them.

The experiment did not go as expected, as it initially appeared that the bacteria were not fatal to Canadian zebra mussels. Mnkantjo switched gears and took an impromptu trip to QUBS to collect more organisms and conduct experiments under more natural conditions.

Mnkantjo spent a week at QUBS conducting fieldwork and mesocosm experiments, which closely mimic the lake environment inside massive 200-litre tanks, alongside Emma Dafoe, ArtSci ’27, an invasive species control technician at QUBS this summer.

Her team canoed across Lake Opinicon and Buck Lake, five kilometres apart, and back, collecting mussels. “I had a wetsuit for the first time. I went canoeing for the first time,” Mnkantjo said.

Another major highlight for Mnkantjo was presenting her research and representing her lab at conferences and talks.

“I’m a Black woman, and I did not see a lot of Black women doing research before coming into this experience,” she said.

Mnkantjo spoke at Civil Forum, a weekly event held by the Civil Engineering Department, a space to engage with research and industry.

“It felt really empowering to be able to bring awareness, whether conscious or not, to other people that Black people do belong in research, and we can provide valuable science,” she said.

The group was able to identify conditions under which the bacteria killed the mussels in Canadian waters, requiring more than 10 times the U.S. limit to achieve similar mortality rates. They are currently working on a protocol to inform Canadian molluscicide, which are specialized pesticides used to kill molluscs, regulations.

A.I. agents get tricked

AI agents going rogue sounds like a sci-fi plotline that has become all too real, but Aaron Su’s, CompSci ’28, research asks — what if we take them at their word? Su, a third-year computing student, studied the trustworthiness of AI agents through an USSRF this summer.

“My project measured whether or not AI coding agents conceal successful indirect prompt injections in their natural language replies to developers,” Su said in an interview with The Journal.

Indirect prompt injections occur when an AI agent encounters a malicious instruction in a file or resource it accesses, rather than in the user’s prompt, Su explained. As companies rush to integrate AI into systems, this poses significant reliability and safety risks.

Su ran 678 trials across nine open-source and proprietary large language models (LLMs), using four different prompt injection vectors, including a docstring — a short string at the top of code — or a file that LLMs read.

In 46.1 per cent of successful prompt injection attacks, the agents failed to disclose the attack in its reply to the developer, he said. Although AI development moves so rapidly that the exact number may have changed, these problems will continue to exist due to the design of LLMs, Su said.

Su was excited by the autonomy he was given to steer his project and the practical skills he gained along the way.

“I did a lot of software engineering in this research role, which was pretty surprising,” he said. Su built a “harness” for his experiment, a program that automates the prompts, runs large numbers of trials, and records how the AI agent behaves.

Su encouraged ���ϳԹ���Դ to increase funding for AI research, saying the high cost of compute can make it difficult for undergraduates to enter the field.

He is eager to continue his research and publish the results and is open to returning to software engineering or another research role in the future.

Smarter Chip Design

AI relies on special computer chips called GPUs which perform trillions of operations in parallel every second, making it possible to get a response in a matter of seconds, explained Matthew Tindale, ArtSci ’27, a fourth-year computer engineering student, in an interview with The Journal.

Tindale received an NSERC USRA to conduct research in the Computing at Extreme Scale Advanced Research (CEASAR) Lab this summer, analyzing how architectural design across several generations of NVIDIA GPUs impacted performance across a wide range of applications.

While AI is the most popular use case, GPUs are used in a variety of fields from climate modeling to cancer research.

“The design focus has shifted to making these AI applications go really fast,” Tindale said in an interview with The Journal. “That means other applications could be potentially left behind. Our results showed that this was exactly what happened.”

Along with Tobias Salamon and Tavin Mistry, both Sci ’26, Tindale helped evaluated three generations of NVIDIA GPUs — the Volta, Ampere, and Hopper — across a range of workloads from scientific computing to industry microbenchmarks which stressed different parts of the chip.

They found that AI applications ran eight to nine times faster, while other applications saw modest performance gains of two to three times. These results can help researchers in different fields choose the right hardware for their workloads and inform future design choices aimed at more balanced, equitable chip performance across use cases.

The era of simply doubling the number of transistors on a computer chip every year, called “Moore’s Law,” is coming to an end due to physical constraints, Tindale explained.

“Transistors have become so small that for it to accurately and consistently carry electrons and conduct a signal, you can’t make them any smaller — otherwise the voltage will leak out and it won’t work as intended,” he said.

Instead of relying on them getting smaller, researchers will have to find different ways to arrange transistors, he said.

“We have to make more conscious decisions of how to use the silicon budget and the available resources to build a chip that’s good for everything.”

The biggest takeaway for him were the relationships he built with his lab mates and supervisor, Ian Karlin. He found it very helpful to bounce ideas off each other and put together his first academic poster.

The team’s poster was accepted into SC26, the world’s largest high-performance computing conference, where Tindale will present in Chicago this November.

Tindale is enrolled in the accelerated master’s program and will continue his research in the CEASAR lab as a graduate student.

From consuming to creating

“Students have a critical role to play to our research ecosystem,” said Klodiana Kolomitro, special advisor on undergraduate research in the Office of the Vice-Principal (Research), who oversees the program, in an interview with The Journal.

Undergraduate research transforms students’ identities from “consumers of knowledge to becoming creators of knowledge,” and helps them develop resilience and critical thinking along the way, she said.

USRA and USSRF awards give students and supervisors a large degree of flexibility to decide what their research project will look like. Kolomitro described it as “a negotiation between the supervisor and the student,” which her office does not influence.

Research is about the process of inquiry and figuring out how to ask good questions which also move the needle to a better, more just society, she said.

“How you go about it, and what tools, approaches, methods, spaces you use will very much depend on the discipline and the disciplinary norms.”

Kolomitro encourages students interested in learning more about research opportunities and labs at ���ϳԹ���Դ to take the free Undergraduate Research Experience online course.