We’re back with another episode exploring undergraduate research! In this episode, students Gabrielle, Hansel, Sarah, Louis, and Jaehyeon share their experiences working on computer science research projects during Summer 2025.
We had some fairly significant issues with our microphone this episode, sorry! The studio did its best to clean up the audio, but you may notice some audio quality issues in this episode.
Chapters coming soon!
Credits
Audio editing done by the University of Toronto's Arts and Science Digital Teaching & Learning Studio, located in the Sidney Smith Building at the St. George Campus.
Mario: Hi, thanks for tuning to In the Loop. I’m Mario Badr.
Diane: And I’m Diane Horton.
Diane: In today’s episode of In the Loop, we interview students who are working on research projects right now, to learn about the huge range of problems they’re tackling and the kind of things they can accomplish in one semester. We’ll also include some tips on how to pursue your own research opportunity.
Mario: Let’s get into it by hearing about the research projects these students are working on. We met them mid-summer, right in the middle of things.
Gabrielle: Hi. I'm Gabrielle. I'm a math and physics student in the Faculty of Arts and Science. In my last year, and I'm currently working on a project Professor Eitan Grinspun, and PhD students, Otman Benchekroun and Mengfei Liu, on a project in computer graphics, specifically physics based animation and geometry processing.
Gabrielle: So my project is in the space of registration, which as a whole it asks, how do we pose or align two shapes together? We are asking how can a 2D or 3D shape that's just a surface, so imagine a balloon or a whole glass figurin, take on the shape or the space taken up by another shape. So we have one shape, and we want to smoosh or stretch or mold that shape until it looks like another shape.
Gabrielle: There are like many different reasons why we would want to align two shapes together. An area that I'm not working in but is outside, of my research is in the medical field. People want to be able to align, like, two organs together. To be able to compare, like, say, maybe a healthy organ with, like, a not so healthy organ.
Gabrielle: In my space, in the computer graphics space. We would like to take the shapes of animals, people or inanimate objects, and even if they look slightly different, we would like to be able to conform them into the position we want using math.
Gabrielle: We've been working on making, our method a little bit faster. And so we've had some of our computations that were done on the CPU, and we moved them to the GPU. And we've done really well with, shapes that are kind of similar or in the same position.
Gabrielle: So, for example, I could have a lion walking and a lion standing at attention. And our project is able to align those two animals really well. But what happens if I have a lion, a lion lying down - there's a sentence - and a lion standing up. How does the computer know what to do? And so I've added, landmarks, to our method so I can, communicate where the head of my lion is and where the feet are supposed to go when I'm aligning two shapes together.
Gabrielle: I've done that as well. We've also been experimenting with ideas that, like, we aren't quite sure about. For example, we experimented, with adding in elastic energy to our method, which states that we have- say we have all of the points on the lion, and we want to make sure that, the distance between two points on the lion stay, like relatively close together.
Gabrielle: And so I implemented that function. And then after showing it, to my lab mates who I'm working with, we kind of decided that we didn't like the look. It kind of made the lion look a little bit like it was dehydrated and kind of like a raisin. So we're still working out, where that's going to fit. But that's part of the process, trying out things that work and trying out things that you weren't quite sure about.
Mario: It’s interesting that Gabrielle’s been working on 3 different aspects of ths project: moving computations onto the GPU, trying out a new technique with landmarks, and introducing elastic energy – there you can see the physics component. One project can involve a lot of different kinds of work.
Diane: I liked hearing about the work not following a straight forward path; things didn’t turn out right away. That's often how research goes. Also, I’ve never heard before of a lion looking like a raisin.
Mario: The next student is also working in computer graphics.
Hansel: So, hello. My name is Hansel. I'm a computer science and statistics double major with an economics minor, heading into my fifth year. I'm currently working on a research project in the field of 3D graphics with Professor Nandita Vijaykumar and her student Steve Rhyner.
Hansel: The research project I'm a part of is in the field of computer graphics, pertaining to neural rendering algorithms, which allow us to manipulate and learn from 3D scenes. I find neural rendering methods pretty exciting, and they're advancing many applications, such as 3D reconstruction from images, generating 3D scenes using a text or image prompt, 3D video calling, and others.
Hansel: Currently, the algorithms and methods surrounding this target specific combinations of hardware and 3D shapes. So our research is in designing, high level framework that is cross-platform, portable, and doesn't require extensive changes to the code if we want to use different algorithms, optimizations and so on.
Mario: Even though Hansel and Gabrielle are both working in graphics, the actual work is quite different.
Diane: Really it is. Hansel told us how he got started and into the project.
Hansel: So coming into this project was definitely a pretty daunting task. And I started out with some bits and pieces of the code base here and there, doing some simple things like restructuring the code base, rewriting certain algorithms to improve speed or compatibility.
Hansel: For the past, while my work has been getting our framework up and running on different hardware, which includes Nvidia, AMD, and Intel GPUs, and writing documentation on how to run that.
Mario: It was a lot of work to just get used to the code. And he started on small tasks before moving into more substantial ones.
Diane: And, as Monty Python would say, “and now for something completely different”.
Sarah: Hello. My name is Sarah. I recently finished my third year studying computer science and cognitive science. This summer I've been taking a research course doing research under Professor Ashton Anderson in Computational Social Science. I am also doing an internship at Relay for working as a full stack developer.
Sarah: So at the highest level, the three big ideas that we were trying to combine are large language models, creativity and emotions. Specifically, there is this notion of the artist-audience relationship where as an artist, when you create art, you have this emotional release. The sense of catharsis and also ownership over this creation that you've made and you are communicating these emotions, these feelings to the audience.
Sarah: And so on the receiver side, as the audience, you experience emotion contagion, in which sometimes you experience what the artist had felt as well, or you might experience something completely different. And our question is when you introduce large language in the mix, when you have LLMs assist artists, how does the artist audience relationship change? And I think the overarching theme here is how does AI change human to human relationships?
Sarah: How does that change how we interact with each other and tell stories? So we are in the process of refining these specific conditions that we want to test. Well, we have more specifically, though, is for playing with the idea of blackout poetry.
Diane: If you haven't encountered this before, blackout poetry involves taking some existing text – it could be from a book or newspaper or anywhere – and removing some of the words so that what is left is a poem.
Sarah: And so we want, our participants to be able to think of a memory, and then they have a blackout poem that they can fill out however they would like. And so we're kind of experimenting with no AI and then having AI at the beginning, or maybe review and revise at the end, and seeing how that affects the end piece that comes out and how the audience, how they perceive those different poems.
Diane: I love that they are exploring how LLMs impact creativity. It's so interesting.
Mario: And I want to write a blackout poem.
Diane: We should try it! But first, let’s hear what Sarah has done so far.
Sarah: We started off obviously with understanding kind of the background, what exists, the literature, what gaps there are potentially. So after reading lots of research papers and being very confused all the time, but slowly figuring out bits and bits, we kind of narrowed down our question a little bit more, not only looking at what could potentially be explored, but also what we are curious about personally, then kind of coming up with certain conditions like how do we turn this big question into a quantifiable experiment that we can run?
Sarah: Then we recently kind of became our own guinea pigs. My partner and I, we wrote a lot of blackout poems, despite none of us being poets in particular. And then now we are kind of moving into the actual implementation phase, looking at Figma designs, coding and then hopefully soon, fingers crossed, we'll be able to deploy it on the internet for people to start playing with.
Mario: So Sarah has spent a lot of time refining the questions she wanted to answer, designing an experiment to run on human subjects, and prototyping a tool for blackout poetry.
Diane: Ultimately the experiment will be run and the data analysed to try to answer their research questions. It’s easy to see how these projects can continue over more than one term. Even with full-time research in the summer, sometimes the work has to continue.
Mario: Our next student is working in computer security.
Louis: Hi, I'm Louis. I'm a fourth year student. I am pursuing a data science specialist and computer science major. This summer I'm working full time with Professor Gururaj Saileshwar in the area of security for AI agents.
Louis: So we are looking at, what people have done on securing AI agents as LLMs are getting better and people have been trying to make them into proper agents that can work on real things, maybe code for you. It's getting a bit harder to think of security in that sense. Like, how are the attacks going to look like?
Louis: And we're trying to look at what people have done in classical software engineering security and applying that principle into AI agents. So the attack that people are most concerned with is indirect prompt injection. How data like malicious data can infect a conversation and cause an AI agent to do things that it's not supposed to.
Louis: So at first we were doing a literature review on what people have done, and we're looking at the adjacent, which is the security in LLMs. And, does it transfer, like does the attacks- do the attacks there transfer properly or do you have more problems because now agents have more things working towards it, like, you have more inputs from user because it takes data from other sources. You have another possible attack vector. So looking at all of those and mocking up little things, trying to experiment. That's pretty much what me and my teammate have done.
Mario: Literature reviews are a big part of research – if you want to do something new, you have to know what’s been done before. So Louis has been working on that and also trying to replicate known attacks or mock up new ones. Quite different from what we’ve heard so far.
Diane: Our final student is Jason (Jaehyeon).
Jaehyeon: Nice to meet you, everyone. My name is Jaehyeon. I'm a fourth year, undergraduate student studying cognitive science, statistics, computer science, and psychology.
Jaehyeon: This summer, I'm working full time with, Professor Suzanne Stevenson and Jai Aggarwal researching in computational linguistics.
Jaehyeon: There are very many different languages in the world and have very different ways of categorizing these their words. So, for example, English categorizes colors into 11 words, but some other languages categorizes with only four. In our project, we're really looking to, see how these types of systems emerge by using information theoretic and probabilistic methods.
Mario: Jason told us what he's done so far.
Jason: I've been working on three main tasks. First working with, cross linguistic data sets, to understand how languages divide up the words. And second, doing a lot of literature reviews in various disciplines like cognitive science, linguistics, computer science, statistics to get a better perspective of how other people studied this, before. And then, third, I've been extending the capabilities of the existing models out there by delving into, as I mentioned before, information theoretic perspective and seeing how it can be applied in a cognitive science model.
Mario: Literature reviews come up again.
Diane: And Jason has also been analyzing linguistic data and working on a new model of language change. These five projects we're talking about today are so diverse in what they’re aiming to produce, and the techniques used.
Mario: Now, since doing research is very different from doing a course – you've probably heard that many times – we asked these students to share with you what it was like to work on a research project. They told us about what made it challenging.
Diane: One theme we heard was that there is a lot of advanced material to learn.
Jaehyeon: So, challenges? I think the, discipline of computational linguistics is not, as, been taught in the undergraduate level, especially with an information theoretic, aspect involved. So there's a lot of learning happening, a lot of new readings, a lot of, foundational, knowledge that I have to gain. And that was a big challenge for me because last summer when I started, I was still finishing my second year, and, it was a big jump from doing an introductory courses into, these- reading long and long papers to understand what's going on in the project.
Louis: I think, the most challenging part is definitely, you know, getting to know what people are doing, like the state of the art. Because it is evolving very quickly, especially in this new secure area for AI agents specifically, like some of the papers we read are published last month, like last week. So it is quite challenging to try to keep up with that and maybe even trying to mock the experiments, because these are research projects and sometimes the code are not there or, you know, the ideas and the results you can't really replicate like right away in an afternoon.
Mario: Related to that, Jason told us he had to question things he thought he already knew.
Jaehyeon: And what's surprising is that there are concepts that I believe I knew very well that turns out to be not true. Or it turns out that there are so many other perspectives of interpreting these, concepts. And there was I think this also goes back to being challenging, but also very surprising, because in an undergraduate level of concepts, mostly if you learn something, that's pretty much it. One big, concept would be about KL divergence. This was- for me, it was first introduced last summer when I started this project. Where back then, I consider we, we consider it, I consider it more as like, how similar to distributions are.
Jaehyeon: And because I was still a work study student, wasn't working full time yet. That was enough for me to, go over and then continue. But now that I'm working as a full time student, full time researcher, and, really pitching in my own ideas, I need to have a much more detailed, perspective and interpretation of what these KL divergences are representing. And, how we can interpret these, values into a cognitive science perspective. And that was a big jump for me as well, from just knowing formulas into interpreting what these values mean.
Diane: Gabrielle told us it was difficult to know when to stop learning.
Gabrielle: In terms of challenging, I would say that knowing when to stop, like reading and learning and wanting to start implementing ideas. And this is kind of also the surprising thing to me too, like how similar it is to school, like how similar my tendencies are because I love to arm myself with knowledge before tackling a problem.
Gabrielle: I know there's some common advice, like for math, where you just jump into the questions and try to answer them, and then only look back when you hit a spot of the problem that you don't understand. I like to read the whole chapter and make sure I know everything perfectly. I do all the examples and then start the problem.
Gabrielle: But of course, in research, you don't necessarily have all the time in the world to do that. So knowing when to stop and knowing when to like say, okay, this is enough. Maybe I'm maybe I need to I need to like, I know, take a step out and you just start now. We need to stop preparing and we need to start doing. That's always something that I need to talk myself into.
Diane: I imagine this feeling is not uncommon. In coursework, if you want to be really successful, you keep going until you master the idea. But in research, the existing knowledge may be jagged and full of holes.
Mario: And because of that, you can’t always look up the information you want.
Louis: So I think the challenging part of this project, and a lot of research in general, is going to areas that people haven't really gone into before, and you really lack a lot of documentation on things and have to experiment on your own to figure things out.
Mario: Another challenge we heard about is honing in on exactly what you want to do.
Sarah: Figuring out the question that I want to ask because there's so many things to be asked. And I'm like, if I have this chance to say something to the scientific community, what is it that I want to say? And so figuring that out has been a challenge and something that we're still working on.
Diane: OMG, did I struggle with that in grad school!
Mario: Okay, that was all about what these students found challenging in research. And I think they found these really productive, positive challenges. Let’s talk about what was fun!
Diane: We kept hearing about the ideas - how exciting it was to be talking with smart people about new ideas on the cutting edge.
Jaehyeon: What's fun has been, really about all the meetings that I've been attending and, all the conversations I was able to have with, the people that I work with because, they offer very novel ideas and the fact that I have the opportunity to pitch in my own ideas into their work is also very exciting and fun to work with.
Louis: It's that you get to see very inventive ideas, this very cutting edge, you know, you don't even conceive of. People trying new things and they're just expressing it, right? Because it doesn't mean that it will work 100%, or even if it's viable in the real world. I don't know, it feels very exciting. Right.
Sarah: My team is so wonderful and I think I didn't anticipate the number of insightful conversations I would have, not only within my team, but also with people that I know in my life outside of research, because I would tell them about it and then they would have all these thoughts to share. And I'm like, that's incredible. I didn't know they had all these thoughts.
Sarah: And so that’s has been really cool. And just this process of throwing mud the wall, seeing what sticks, taking the time to ask questions, not being very pressured to have something tangible every single time, we meet with my teammates. I think this freedom has been so much fun.
Diane: The breakthrough moments are pretty great too.
Hansel: The fun part is that once you get a working, it feels pretty rewarding.
Diane: Mario, do you remember any particular exciting moments in your research?
Mario: Yeah, absolutely. I think I did a lot of modeling in my graduate school research. And the one that, well, any, any of them actually. But there's, there's always that moment where you realize, oh, that's what I haven't considered in my model. And once you kinda like have that moment, you like add it to your model and then everything starts to work. It's that's, that's probably the exciting part.
Diane: That sounds exhilarating. For me, I think about the anticipation I felt when I was sitting at the R console about to run a statistical test to find out whether an educational intervention actually produced a significant difference in the outcomes we were measuring. And it just, it was so exciting to know that I was going to find this out and that it would be applied to my teaching. Really wonderful.
Mario: Your exciting moment sounds so much more sophisticated than mine.
Diane: I don't think so.
Mario: By now, listeners, we hope you are excited to get involved in research if you haven’t already.
Diane: Our guests told us the story of how they found their research opportunity. Some research opportunities in the Department of Computer Science at the University of Toronto are listed publicly and students can simply apply to them. (We’ll provide details at the end of the show.) Several of our guests took this route.
Louis: So I think there's a Google form that you just fill in with your personal information maybe. I think there's a resume CV part and a cover letter. You can submit to multiple projects there, I believe, and rank them based on your preference. So that's what I did.
Hansel: So, I found this project and supervisor through applying to CSC 494. When I applied to this course, there was a list of projects that you could apply to. And among you can rank for of your own interest. And I happened to get into the one that I currently am.
Sarah: Yeah. So when I first came into university, research was not something that was really on top of my mind. But then I started noticing that some of my friends were getting into research, and when I had conversations with them, I was like, this is incredible. I want to be a part of this as well. So I reached out to one of my friends about her research experience, and she said, I'm doing a CSC49 and you just you watch out for when the announcement comes out for the postings and you apply.
Sarahh: And she was like, just do it. You might be a little bit nervous. Just apply. It's going to be good. So for me, once I got the announcement, I personally filtered for human computer interaction related projects. So you can filter for whatever interest you have.
Sarah: For me, it was HCI. And then I was bring into the projects and I was like, when I saw LLMs and creativity, like a little light bulb went off in my head. And so that's kind of how I chose to apply. So for me personally, the project came before the professor, but I know some people, they are big fans of a professor and the work that they do, and it's kind of the other way around.
Mario: Gabrielle took the approach of talking directly to a professor – one she didn’t even know.
Gabrielle: So I can speak, to how I met, my supervisor, Eitan Greenspun. So way, way back, in 2023. I went on the Department of Computer Science website. Clicked on the research tab, clicked on computer graphics. And I was redirected to the Dynamic Graphics Project website. I clicked on people. And so I clicked on David Levine's face, Alec Jacobson and Karan Singh’s. And so I looked at, Dave Levine's website and he had like a tagline mentioned that he loved mentioning that he worked in physics based animation and he loved numerical simulation. And then one day in between classes, I had some time and I worked up some courage and I said, okay, I'm just going to go to his office.
Gabrielle: And, I waited till he was speaking, finished speaking with the student in his office hour. And I went and I introduced myself, and he invited me to the weekly DGP lab meetings. And I've been going ever since. And then at one of those meetings, Eitan Greenspun is one of the co-chairs of the meeting or one of the professors that regularly attends the meetings, and he asked me, if I wanted to do research. And I remember him looking at me and then in my head, I was just like, oh, wait, no, I'm not ready. No no, no. And then my outside, my my mouth said yes. And so that was how I got into research.
Diane: I love that Gabrielle told us it was scary to approach this prof, and that she didn’t feel she was ready to do research, because I think these are very common feelings. And I love that she did these things anyway.
Mario: Yea and it's very inspiring, and look how well it turned out!
Diane: What’s our favourite saying?
Mario & Diane: “Talk to your profs!”
Diane: Okay, let’s wrap up with some final tips for students who want to get involved in research. First: Don’t let shyness stop you from getting involved in research.
Hansel: So even for students that are a bit shy and find it quite intimidating to approach professors either in person or by cold emailing, there's still other opportunities, like your research course in CSC494, where you have a list of proposals that you can go through and try to see which ones pique your interest, or you have past experiences in.
Mario: Second, if you’re going to talk to a prof, it’s very helpful to prepare first.
Gabrielle: Have good and specific questions ready to go. Or I would say have specific questions ready to go. So if my, courage brave plan, didn't work, with, going to, like, Dave's office, I the plan would have been to find a research paper, find any research paper, and find one thing that I could connect to a class that I've done previously.
Gabrielle: And then start working on that part and then have a question and say, hey, this is similar to what I think I've done in class. Here's what I think you're saying in this piece of your research paper. Do I have that correct? So instead of jumping in with, hey, can I do research? I wanted to open up with a question. Especially specifically at the time, it was because I wasn't sure if I was ready for research yet, but I knew I could start a conversation.
Diane: Third, be open.
Sarah: In terms of finding a project, I would encourage you to be open minded to exploring different things that you might have never heard about, but also at the same time coming with a certain idea of the things that you're interested in.
Sarah: So finding a balance, if that makes sense. So you are able to at least know a little bit of what you're looking for. But also being like, I'm excited to see what's out there.
Mario: Fourth, consider the environment.
Sarah: And then the second piece of advice I would give is that in addition to the specific, perhaps academic field that you're looking into, I think it's also important to consider the research environment that you want to be surrounded by.
Sarah: A friend recently mentioned to me how research labs are in and of themselves, also in organization. And so there's always a culture that surrounds it. And so making sure that you have an idea of sort of the culture that you want, maybe it would be good to talk to someone who's already in the lab and seeing what that's like, so that you can be excited not only about the work specifically, but also about the people in the community that makes up the lab.
Diane: Fifth, be aware that you’re going to be working in a new way. Research is not like a course.
Louis: If you're interested in research, I would say just try it and see if it's good for you, right? Because it's very open ended. It's different than what you might expect. Especially if you're not working under a project that someone else started. If it's your own project, it's just you are thinking about new things. It's all your ideas.
Louis: But it's also because of that you are, you know, treading into deep waters yourself. You just have a supervisor there. You can ask, what might work, what might not work. You explain to them your concerns and they will help you and guide you a little bit. But it's free form. You're doing everything yourself, and it might be very rewarding to you, but you might not like that as well.
Mario: And finally, if you can make yourself available over more than one term, it can allow you to do bigger things.
Jaehyeon: One thing that the people that I work with really liked about me when I first joined the lab was how well I can express my ideas. Being a good communicator is, something we can't, emphasize enough because it is such an important, aspect, and a lot of students might not think it's as important.
Jaehyeon: Most students might think that doing well on schools would be enough, but I think a lot of, professors, a lot of, grad students also look for how well you can communicate your ideas. And I think that was like a number one thing you should really focus on. Another thing that I think is very important is, your time commitment.
Jaehyeon: Yes. As a, work study student, you might be capped at working, ten hours, 20 hours per week. I think it's ten hours. Yeah. Yeah, ten hours a week. But it's not how much how many hours you can work per week. It's more of how long you can work with, with the team. Because, working with them for just the summer and leaving isn't as beneficial for the- in the lab's perspective, because for me, it took four months to just learn what was going on.
Jaehyeon: And, from there, I could actually do some, meaningful work. So, I think time commitment is also very important. And lastly, most students should really, look into other departments as well because, a lot of departments are also looking for great, computer science students, and also for non-computer science students, the CS department is also looking for a lot of, interdisciplinary students as well, just like myself.
Jaehyeon: So looking into these, other labs, other opportunities, should really open up more, opportunities that you can, look for.
Mario: What a great look at how students are doing research over the summer at the University of Toronto. If you want to prepare yourself for research, pursue that first research experience, or are thinking about grad school, then check out our show notes.
Diane: UofT CS students, you are studying at one of the world’s best CS departments, and there are tons of opportunities for you to do research as an undergrad. We hope this episode inspired you to think about it, and showed you some concrete ways to get started.
Mario: Alright, let's wrap this up. My name is Mario Badr.
Diane: And I am Diane Horton.
Mario & Diane: And you are In the Loop.