Chris Yip  0:01  
Welcome to Tell Me More: Coffee with Chris Yip, the official podcast of the Faculty of Applied Science and Engineering at the University of Toronto.

Each month, I sit down with someone from our vibrant global community to talk about what places them at the heart of designing bold solutions for a better world. You'll meet students, professors and alumni who are making a difference across a range of fields, including some where you may not expect to find them. This month, more than 1000 U of T Engineering undergraduate students will walk across the stage at convocation, and among them will be Shrey Jain, who is completing his degree in engineering science with a major in machine intelligence. Over the past few years, Shrey has worked on a number of high profile projects related to AI and its implications for our society. Work that he'll be continuing after graduation as an applied scientist at Microsoft. Shrey, welcome to the podcast.

Shrey Jain  1:00  
Thank you, Dean Yip. I'm excited for this.

Chris Yip  1:03  
We've chatted a lot over the past, I can say the duration of your undergrad degree in a whole bunch of different contexts. I'm really looking forward to getting all the listeners of the podcast a little about...all this cool stuff you've been doing. We always start with this, the classic question, you know, when did you realize engineering was for you? And of course, the most important question, why did you pick U of T Engineering?

Shrey Jain  1:28  
I never thought I was going to go into engineering or anything technical as a kid. I grew up in a family where the predominant area of focus was medicine, and no one ever dipped their toes into anything technical. It wasn't until like high school where I went to this summer program, it was called Shad, where I got exposed to a lot of kids getting really excited about going and applying to engineering school and I like, didn't even know what it meant. I didn't know the difference between civil, mechanical or electrical, and like all the different types of engineering, but that summer, I got exposed to working out in the University of Saskatchewan on the synchrotron there and got to work with a lot of students who really opened my eyes to like what it could be like to do engineering and I didn't even know what that meant still, at the time. That was probably like the flipping point and then I started to go to a bunch of hackathons in high school, and started to get more into the computer science, machine learning space ish, like at that time, again, wasn't really sure what machine learning even was, kind of just thought it was cool. But then, yeah, started to apply to different engineering programs in Canada and there's a lot of top schools in Canada, all of which that have their differentiating factors. And at the end of the day, I was looking at U of T in particular, mostly because of the engineering science program, I would say, in particular, at the time, what I thought would be the best decision for me was a program where I'd feel really confident coming out in my ability to teach myself anything. I do think I did believe and still do believe that the school program itself will teach you maybe the fundamentals and as well as maybe some really critical concepts but in the workplace, it may be outdated and we're seeing like a bit of that already with maybe the courses I took even last semester and what I'm going to be working on in the fall, just because of the pace that things move at, especially in AI. I thought that okay, if I go through this very challenging program, and maybe it's like a meme now in EngSci in particular, that everyone likes to do this challenging thing, that I would come out with the skills and confidence to be able to teach myself anything and I think, in retrospect, any program at U of T would have done that, but for me at the time, it was EngSci.

Chris Yip  3:37  
I was gonna call you out, you did say that everybody in your family went through medicine, which was not a technical field. I will qualify that for listeners, I think there is some technical stuff in medicine. [laughing]

Shrey Jain  3:49  
Yes, this is true. My dad is a surgeon and there is a lot of technicalities to that job as well but a different type of technicality in the way that they still need help fixing Wi-Fi.

Chris Yip  4:00  
Well, okay, fair enough but there is also a there's a pretty cool intersection, obviously, between engineering and medicine as well and I think that's a really neat area. So you're in the machine intelligence option. That's one option, I think was it that it started when you were coming in first year or just around then, right? Sort of just been kind of born in as an option inside inside? So what sparked your interest in that kind of AI machine learning space?

Shrey Jain  4:25  
Having been in, again, like exposed to the hackathon community I think in high school is when I first started to like play with OpenCV which is like image detection software for like object detection and again, it was just something where when you see it and you're like using it, it's amazing that a computer and something that you could program could work. In high school, I was fortunate enough to have an instructor who kind of took all the students, there's like eight of us who are like extremely passionate about software and computer like machine learning, getting exposed to that even in a very elementary way. I think at the time, and it's crazy to see where the progression has come because it's been like six, five or six years since then, is incredible but at the time was like, wow, I think that this is gonna have a really big impact, I don't really know what it is but more importantly, I just thought it was cool. And I was like, if I'm going to do that, I think I want to do it in like choose that major in EngSci. I also at the time in high school had reached out to who now is my thesis advisor, Professor Stark Draper, and was like, a fanboy of a lot of the work he was doing and the fact that I knew he designed the major was also really, really exciting. And I've also maybe it will come back to you later, I've come full circle with him as an advisor to me through undergrad as well and so that was the main reason why I chose it. 

Chris Yip  5:45  
You lucked out, right? You lucked out by being that cohort that came in first year, and landed with this thing called COVID, a pandemic hit during first year for you, right? And first off, let's start like, what was it like for you and your fellow students when you're like, hey, what the heck?

Shrey Jain  6:04  
Yeah, it was crazy and I guess there's a small snippet of this story that I've shared with some people, which is, again, I guess my family was pretty well read on what was going on in the pandemic, globally. And at the time, I was living at one of the residences that a lot of engineers live at, Chestnut residence at U of T. My dad told me to move out two weeks before, there was a declared pandemic, which to a lot of people was like, "What? Is this kid's family paranoid?Like, is he crazy? But no, I like I moved out early because my dad was like, we're not going to wait for the elevators in two weeks, when they make this decision. I know what's going to happen. And at the time, like if I'm being completely honest, I was still adjusting to transitioning to this very difficult program in first year, my friend group size was very small, because I'm still working out the kinks of adjusting to like a university lifestyle, and like managing academics, athletics, along with social life so I haven't really had that time to even establish like my U of T Engineering friend group yet, and all of a sudden, you're home and it's like, I'm living with my parents for two years, because I have immunocompromised people in my household and you can't leave. And at the time, it was just it was crazy, I think, a lot of fear to amongst students, but also excitement because we didn't have to write finals in the way that we normally did, I guess.

Chris Yip  7:22  
We said this to multiple people sort of full props to all students, and everybody in the community for being able to to keep the smooth ship U of T Engineering or University of Toronto, or actually the world rolling forward over this time period. It was interesting, because it was an opportunity for so many different things to emerge, right? I mean, we found out at the university is like, hey, actually, we can teach on, we don't like teaching online and I think you guys would agree, that's probably not the best thing to do. But we figured out how to do it in ways that can be useful actually, even now. There's some good things and some bad things, let's roll some good things in as we go forward out of the pandemic, and but at the same time, other opportunities appeared for people who are kind of entrepreneurial, or realize there was a need to get information out, right? And to help kind of with the pandemic, right? You and your friends kind of did something a little different. You want to tell us, tell us a little bit of what flatten.ca was all about?

Shrey Jain  8:16  
Yeah, for sure. So I moved out at the time, I was saying and was kind of at the time, I was just starting my new position as an undergraduate researcher at the Vector Institute of Technology, which is one of the bigger like AI labs around U of T. And started under my advisor at the time was Marzyeh Ghassemi, and she was working on health in AI related research. And we had a lot of people in the in the lab, who were looking at COVID and like what all of this meant, and how we could start doing modeling on COVID data. But in Canada, there was no modeling to be done because we didn't have data and I want to acknowledge two other key people to this project, who are also both U of T Engineering students here. One is Martin Staadecker, who is also an EngSci 2T3, and Arthur Allshire, who's also an EngSci 2T3. With them, and along with some other students who contributed in different capacities, we wanted to find a way to get data so that people could start modeling and having some sense of like, where COVID was and like, again, at the time, I'm a first year I have no exposure to like how to do surveying properly, how to manage public health data properly, or let alone healthcare, very sensitive data properly. What we ended up putting out as like an MVP, was this survey with U of T students around what are your symptoms? Like, do you have a cough, fever, cold and we didn't do this in any formal way. Like we didn't ask for permission from the school, can we send out this survey? At the time we were like, let's just see what people's data can show. It was really interesting because like we started getting some people who actually were symptomatic of COVID that we then just put on a map, like a heat map where we just put red dots on a map of The Toronto area and made that map a little bit more red than other areas, there was like a super easy concept, it was like, create the most easy visual for people to follow. And I think a lot of people like, we didn't know why at the time believed in it and it was because the we were collecting more data over the course of those two weeks than the federal Canadian government was. Like we had over a million people contributing their data but this was only due to also not only the fellow contributors that I mentioned, but also from the advisory we had. One key advisor that I think maybe a lot of the outcomes we had would not have been possible was with the advisory of Dr. Geoffrey Hinton, who is one of the more world-renowned AI researchers who helped not only introduce us to a lot of key influential people in Canada, whether it be like for funding purposes or otherwise, but also just to like, make it very clear on what should happen from a data perspective here, and also having the support financially from various different partners to be able to fund six engineering students at U of T to work on this full time for an internship when their internships got canceled. And so over the course of that summer, we got to be the primary data collector for the City of Montreal, we got to work with the Chief Medical Officer in all of Canada, along with like various different other people and learned a lot about one, data collection surveying as well as modeling that data. And then we got to do a whole range of other projects but the maybe the one brief one I'll mention is we then spent another consecutive year working in Somalia remotely, working with collaborators at the United Nations, the WHO, along with the government in Somalia, to replicate the process we had in Canada, because once we scaled up like antigen and other types of nasal testing, we no longer needed this, like, do you have a fever or cough? We needed more like hardcore data. But in Somalia, they never scaled up that testing and so that tool became really useful in that interim period before testing got scaled up there.

Chris Yip  12:01  
Then to think that was, that was just your first year. Like, wait, we got to, we got three years to talk about. But you continue to kind of do stuff at that AI and medicine interface so it's kind of cycling back to the fact that you still have to give props to your parents being in the medical space. But what other projects were you involved in?

Shrey Jain  12:20  
Yeah, so I guess there's a couple of like side projects that I was helping out with. So at the time I moved labs, like I moved from Vector Institute to MIT, I moved to the CSAIL lab there, mostly because my advisor moved labs like Marzyeh moved to become faculty at MIT at the time and so I was working there and then I actually ended up again, switching labs to the MIT and Harvard Critical Data Lab, where I got to learn a bit more on privacy techniques on how we protect privacy on releasing healthcare data and like, how do we unlock healthcare data generally? It's like this really big problem in medical AI research. A lot of my mentors and advisors were telling me that I should go and try it to see what it was like to work out at a big company working on health and AI, and see what types of contributions I can make as a pm or a technical product manager there. And so that's when I stopped doing more academic lab research and started doing research and product managing work at Microsoft and worked on the health care in AI team there, where it was a totally different problem where it's like, we have too much data and we need to make sure we handle it responsibly with our partners and move from this like very small environment in the academic realm to like this big tech corporation of like 200,000 people. For context, Microsoft invested $20 billion into healthcare and AI at the time when I joined that lab, which was more funding than my brain could process. I got to work on a lot of really cool projects, I got to work on projects related to how to build efficient data pipelines for healthcare data, so that if a patient in Germany moves to the United States, their allergies are going to show up on the dashboard in the United States so that someone doesn't treat them with the wrong medicine or learning about how to use the blockchain for healthcare data. And got to learn a lot about the big tech ecosystem as well and still there so maybe it's something, something clicked there for me as well.

Chris Yip  14:18  
You're working on these challenges tied to this sort of what's known as generative AI, right? There's GPT-4 and DALLE which create images and text based on on large language models, right? And can you talk about some of the challenges these things raised?

Shrey Jain  14:33  
There's a lot of challenges here. Maybe I'll like scope it on specifically the ones that I'm working on. So yeah, there's all these different types of generative AI tools or large foundation models, whatever language you want to use here, of which is increasingly powerful and seemingly looking like you're a human or creating very realistic text, images and video in a multimodal way. And there's a lot of fears with what that means for the ways in which we as humans interact with one another. Because if I am no longer able to distinguish that I'm actually talking to Dean Yip on this call, how does that affect my relationship with you in the physical world? Or how does that affect my relationship with you in my imagined relationship with you or others, and there's a lot of like very real examples of this already. People are taking snippets of people's TikTok videos or Instagram videos, and synthesizing their voice to then scam their parents or their grandparents to say, Hey, Mom, I forgot my social security number, do you remember what the last three digits were? And then they send them the social security number. Next thing, you know, they have access to a bank account, and you lost maybe $21,000. There's like that issue. But then there's like the other issue just generally, which is what happens to public trust? What happens to the ways in which misinformation spreads? What happens to the ways in which students interact in the classroom? Like, I remember, throughout my entire last month of school, I was in the library and if I walked down a given aisle in any library, and this is not something strictly at U of T, it's at schools across the world from what I'm hearing from students, is there's a split screen between students work and Chat GPT. And it's like, what does that mean for academics? And what does that mean for instructors? And like, what does work look like? And it's my work focuses on how do we determine whether or not someone is in fact a human behind a given message? And how do we ensure that we can protect someone's identity from never being able to become synthesized or reproduce in a given way? Microsoft researchers, which is where I work, I work under the security and cryptography team. And we have various different techniques that we try to apply to the ways in which we train models, the ways in which we transmit messages, the ways in which we sign messages to strengthen those guarantees. But more recently put out this paper, which was very well received by the academic community and some of the top AI labs, like open AI and others, was related to what I called Plural Publics. Plural Publics was this paper that we designed this system and scheme for how we can communicate in this new world where generative AI takes over. Like how do we distinguish between each other? And Plural Publics, which is related to like basically creating the most trustworthy group chats ever online. The short of it is like the internet throughout history has never taken privacy or security as seriously as maybe we need to now. And what I mean by that is like, we've never really had our hygiene up to date with what maybe we should have been doing all along, which is like, Yeah, I know, I should change my passwords every two to three months but I don't, or like I know, I probably shouldn't, like write my password on a sheet of paper, and keep it in some drawer loose leaf there but I do. And it's like, these have always been nice to haves but in this new reality where your identity is at risk of being forged, and the ones around you are at risk of losing part of their identity as well, it maybe goes from this like nice to have to like a very critical necessity. And under those assumptions that we're making, we're trying to build the tech that enables people to protect that. And also support them in doing it not just under the Microsoft umbrella, but doing it with our collaborators in open source again, and doing it with people in a way that people can all contribute and deliberatively decide on how we go about building it. And so this is where cryptography plays a big role in it and that's where a lot of the focus of our work our work is where it's like, how do we apply cryptographic techniques to support us basically, keeping these dictionaries that we have with our friends private, secure, in a way that no one else can really tamper. And I think, although there has been a lot of this, like beauty that we have had with the internet and this like instantaneous connection to one another, because that security has never been like a focus but now it may become a focus, it may even become a better internet than we've never had in the sense that like, now this I'm not afraid to post online because who I'm posting to online is like a group of people where I have shared trust, or social norms may change in a very healthy way if we can get people to work on these problems collaboratively. And, like there's also a lot of optimism, I guess, is what I'm trying to say here.

Chris Yip  19:26  
Let's shift off of this to something more fun. So you're you're a triathlete, right? What does that like? And how does it help? How did it help you stay balanced, I mean, throughout all the cool stuff that you did during your undergrad?

Shrey Jain  19:37  
I'm happy we're actually asking this question now because for the past year, I haven't really been doing triathlon. And I've realized the value it had on my life when not doing it because everyone has a different like way of like exerting energy or distressing and for me growing up as a kid, I never did like technical clubs like I never did robotics, I never did like those math competitions in middle school or anything. My time was completely filled with sports. I remember because I'd wake up in the morning, my mom would take me to swimming, I'd go to school, and then I'd have swimming after school, and then if I didn't dry my hair, I'd get almost sick because I had hockey after swimming. Like my whole day was like sports and all the way through high school, the same thing. But I think it's like been really helpful for me as an individual, because that is my way of like, keeping everything in check, whether it be like my sleep, my eating, my like, my social habits. I met some of my best friends through sport. I try to find that balance. Obviously, I think that is a weakness of mine with like, some things have to give in some areas where it was like, on Friday night, maybe you don't go out because you have a Saturday morning practice. And that may have incurred some costs socially but at the same time, it kept me in check with regards to like my food and nutrition. And I've become obsessed with monitoring myself and my body in the sense where it's like I'm wearing an aura ring, I try to track all my data through my Apple watch and I think taking that very like quantitative like engineering approach to my health has really helped me honestly with like, understanding how it like how I feel and I feel like I'm pretty in tune with like, knowing when I'm not in a good health place and am and triathlon lets me just zone out, like completely from everything. And I'm sure you can relate to this feeling with a runner's high. It's a feeling that you don't get until you do it consistently enough to know that every time you go to or I should say, 1/3 of the time you go to a workout, you will likely get that feeling. Yeah, this last year of college, I stopped doing triathlon because I told myself, I wanted to go harder with school and I was working part time and all these other things and I regret it now having done that, because I think I didn't make much more progress, honestly, professionally. I ended up just making more time to do the things that I had and my sleep was lacking and I didn't eat as clean as I normally do. And so now that I've graduated, I like feel like I tried both ways and I think I learned that triathlon for me, or like intense sport keeps me in the best version of myself. And like, it may be sport for some people but for others, it may be other activities. But yeah, so that's that really helped me through undergrad.

Chris Yip  22:23  
No, for sure. A cool lesson learned. So hey, June is coming up, right? What's the plan? And I guess for the listeners, what's, what's the big lesson that you're coming away from U of T Engineering with?

Shrey Jain  22:37  
Yeah, so the plan is like, try to work on things that I feel really excited and passionate about. And for at least the next year, that's going to be working on this problem of like, hopefully preventing some of the most existential problems to human communication that I think we've ever had, ever, and doing it at a place where I think I can have extreme impact because of the visibility and work and mentorship that I do have. But if that starts to deter, and not become something I'm passionate about, then I will be like moving and figuring out what that is. There's a reason why things are cliche. Like it's cliche, because it's like a reality that a lot of human beings who are social atoms have realized which is like, do what you're passionate about and that's actually like a reality, because at least I am feeling that emotion where it's like, I don't feel bad about going to work because I'm excited about it all day. The learning over four years, it's so fresh still. I'm still processing a lot, obviously but my initial reaction is like, there's like some hard realities that you have to learn, which is like, be good at detecting the fluff in terms of both work but also like with yourself, like what is actually helping you progress as an individual versus what do you think is a social norm that you don't feel good about. And two, is like, learning how to teach yourself is probably the most important skill you'll learn in undergrad, from me at least, because I come out with a Bachelors of Engineering is in my degree, I think or Bachelors of Applied Science I think is the title. Another way of framing it, at least in the ML major could be like a Bachelor of the History of ML, because at the end of the day, what's happening even a month ago is no longer relevant today in a lot of ML field. But what was relevant throughout my four years was like collaborating with human beings who also wanted to be like solving some very pressing problems, and do so in a way where even if it wasn't taught in the textbook, we worked with our professors or the environment in the school to support us to go and learn and teach it and I think that's a skill that I'm going to try to instill whether it be at the workplace or eventually maybe graduate school.

Chris Yip  24:53  
So last question, you have a younger brother coming up, potentially coming through U of T, right? What are you going to tell them? You can say, Hey, don't do this.

Shrey Jain  25:03  
Well, I guess for the podcast listeners, like I am very clear on never giving advice to people I don't know. So for my brother because I know him very well, I feel pretty good about giving him advice as an individual who I think I have that context. So take that within consideration when I speak here. I know that type of individual he is on this like spectrum of like academically obsessed, socially obsessed or distracted by the like Silicon Valley tech culture axes that I think many engineering computer science students fit on and I told him, just be aware of when you're on different areas of this scale. You might become obsessed with getting a 4.0, or whatever GPA level you want to get, you may become obsessed with like, going out in Toronto in the city, that's the craziest city in Canada, you also may become obsessed with getting into the tech culture weeds of Silicon Valley blogs, or whatever it may be, and you just need to again, as we've kind of been mentioning, in this podcast, find that balance and prioritize people over everything when it comes to your experience, because I look at what I'm coming out with and what I'm most excited about and proud of. I honestly couldn't care less about my grades, not that I'm like extremely proud of them. I couldn't really care less as much about like, any like labeled accolades but I'm proud of like, really two things. It's like the contributions I had, and the shared experience I had with people because at the end of the day, those shared experiences is what I'm going to look back on and remember, and when I go online and FaceTime a friend, if I want to authenticate them in the digital world, I can talk about all of our secrets and shared memories we had, whether it was staying up late in Myhal, or like doing these different things that I want him to try to focus on also having because at the end of the day, if you wanted to just go and learn about thermodynamics, or whatever it was, you could open up a textbook, but you're coming to a university to do it with people and like really, really try to enjoy that experience.

Chris Yip  27:06  
Very fine advice and actually a really, a terrific perspective. It's the people you met, it's the experiences you shared, it's the opportunities that you took advantage of while you were here that really define your success. Really, really, thank you so much for taking the time today.

Shrey Jain  27:25  
Yeah, and thank you, Dean Yip. And for like all the support you've had basically like since first year. It's crazy that it's come full circle to this moment now and I can't express how appreciative I am, the teams I've worked on that have appreciated your endorsement or support along with like, all the other advisors I've had at U of T through yourself and others. So thank you.

Chris Yip  27:45  
Yeah, anytime. Looking forward to shaking your hand as you crossed the stage at convocation. It's gonna be a ton of fun.

Shrey Jain  27:52  
Yeah, thank you.

Chris Yip  27:55  
Thanks again for listening to Coffee with Chris Yip. If you want to catch up on past episodes, or make sure that you don't miss the next one. Please subscribe. We're on Apple Podcasts, Spotify, and more. Just look for coffee with Chris Yip. You can also check out @uoftengineering on Twitter, Facebook, Instagram and LinkedIn for more stories about how our community is building a better world. And finally, if you'd be inspired to join us, we'd love to welcome you. Whether you're thinking of taking a degree or working with us on our research project, you can find us online at engineering.utoronto.ca or you can visit our beautiful campus in Toronto, Ontario, Canada. I hope I can join you for coffee soon.

Transcribed by https://otter.ai