 Good morning everyone We'll go ahead and get started. So I'm Mark Lundstrom. I'm interim dean of engineering And if we're fortunate the next dean of engineering will be with us to close out this morning's session So that'll be neat. So Let me first of all introduce our four associate professors that we'll be hearing from Dr. Harsha Honapa from industrial engineering Harsha is here right And joseph raspoli from biomedical engineering Ivan Kristof from mechanical engineering Ivan and christopher goldenstein from also from mechanical engineering. All right, so it's great to be With you all today. I really look forward to these sessions So you've all achieved a recent milestone And what this means is that your colleagues have confidence in you to make the right decisions about Where you can have the most impact on Your students and your technical communities and beyond the campus and that we look forward to working with you for for many years So, you know when you achieve a milestone like this, it's useful to pause and think about how did I get here and what's next for me And that's what this session is all about You know, there are a lot of transitions and changes that are going on here at purdue So we're at the end of the mitch daniels decade And at the beginning of what I hope will be the mong chang decade And we have a transition in deans of engineering going on So it's useful just to stop and think about where we are and where we're headed as a university and as a college And I think it's about Scale and excellence and impact so scale is about our land grant mission to provide The best education we have to as many students as we can excellence is about Providing that education at a level that's second to none But also at a level of our research that is second to none and you know excellence at scale That's what we're all about and that's that's really a big challenge And impact, you know it goes along with excellence, but it's having an impact making the world a different place really making a difference in Our technical communities and beyond our technical communities in society as well So with that, I will Turn the floor over to dr. Yongjun sun to introduce our first speaker Thank thank you dean runs from It's my great pleasure to introduce doctor hasha honapa who joined Purdue 2015 after completing his phd at usc electrical engineering Hasha also had an industry positions such as a fecal Indian institute of science and bad labs, okay as well And hasha we will hear a lot more about he is actually technical work, but he is the expert on mathematical modeling and analysis of stochastic systems Core centers and uber rider, you know the shared riding systems And cloud computing services and hospitals and you will see the waiting line everywhere And hasha helped to reduce those waiting line in our every day You know their life And hasha received the several prestigious award including nsf korea award and also he received the best outstanding graduate mentor award from college of engineering And then he's a phd dissertation Received the best phd dissertation in queuing theory that is a provide by annually Provide by european conference on queuing theory. That's a very very prestigious one. Okay, so in terms of the educational and you know service contributions hasha also actually excels and So hasha is currently faculty advisor for infom's student chapter at Purdue and had Uh, advised several senior design project and I really like this one So hasha has been working with our undergraduate advisors to develop a predictive model of student performance So that we can provide the uh, uh, you know the enabling data-driven student advising at scale Okay, so and then uh Several months ago hasha proposed this idea that he want to use a ai To you know the try in his course to write the much of the you know code You know programming code from scratch now everyone knows that's a chat gpt And hasha is a very interested in innovating this Education platform and technology in classes. Let's welcome dr. Honapa Well, thank you everyone for being here It's a real pleasure to be talking to you So I thought about what to talk about and eventually I decided that I talk Sort of a parable about my journey and how I ended up here talking to you So I should start by thanking my wife ashwini. Who is there in that corner over there? I'm here standing here Thanks to her encouragement and her patience with my ranting You know with my work and you know the progress I'm making or not making um, you know, we built a very intricate and complex and interesting life I would say sort of like that eiffel tower that we built over the christmas break And I should also point out the character who's very non-plussed here. That's yuki our dog She doesn't like hugs, but she's a lot of fun I should point out that I am sort of in the dog house with ashwini a little bit because She runs a podcast called through the corporate glass where she explores people's career choices And she's asked me repeatedly to come on this podcast and talk about my career choices And I refuse saying that I don't like navel gazing And yet here I am navel gazing. So ashwini, I'll be on the podcast at some point. You've done 67 I don't I don't think I can be the 68, but maybe the 75th or something. I don't know we'll see Um, all right, so before I get into my story, I should point out that my success is sort of premised on You know several people Most importantly my students some of whom are here My graduate students at the top. I've worked with a whole bunch of them. Some of them are in i.e. Some of them are in math some in stats um The second row are some undergraduates that I work with And I continue to work with I always love working with undergraduate students because you know, they bring enthusiasm and joy to their work. They love to learn And of course mentors and colleagues. I will mention some of them as I go along So this is usually the slide that I flash when I try to explain what I work on. I think sound did a good job I don't think my work is going to help reduce waiting times in the short run, but maybe in the long run but I mostly work with probabilistic tools. I analyze stochastic systems increasingly machine learning and You know optimization methods Okay, so let me get into My journey and how I ended up here. So it's sort of a very pathetic decade. I would say before I got to Purdue So my story starts in Bangalore or Bengaluru as it's called today in our local language In fact, my story starts with a failure like I desperately wanted to get into the Indian Institute of Technology Which if you know anything about India and technology, I mean, that's the place you want to go and I didn't make it instead I I didn't pass the exam. So I had to attend a local engineering college as we call it in India and I studied electrical Engineering well electronics and communications engineering to be more precise And I wouldn't claim to be at the top of my class. I I would mean I was Really interested in the work, but I never was at the top But along the way I had the opportunity to work at the Indian Institute of Science As part of my senior thesis. So I spent a lot of time with my friend Just the two of us my Classmate where we worked on protocols for sensor networks And so this was at a time when sensor networks were a big deal And we were trying to analyze some existing sensor network protocols And as part of that we sort of decided that okay, you know, it's really hard to control These little devices on these networks. And so we needed like a sort of probabilistic algorithm And we just couldn't come up with one because we didn't have the mathematical Capability to do it. We didn't understand probabilities to castics well enough And both of us decided that okay, we have to go to grad school And that's how I ended up at USC University of Southern California In 2004 immediately after I graduated I decided I'd go there and study communications protocols and some information theory And thinking that I would somehow become a communications engineer The very first class that I took at USC was on probability theory and that really opened my eyes That's when I sort of understood that. Okay. I need to be in probability I need to understand stochastics and I want to be a more of a probabilist than anything else more than an engineer And I also got really interested in machine learning at that time And so I was very privileged to work at the information sciences institute at USC In the artificial intelligence division On you know, basically some you know machine learning algorithm for Chunking up a sequence of characters into words. We were trying to decide, you know, where do words emerge from so sort of like a proto language model if you like So at the end of this period like 2006, I sort of had a choice to make So this is the first choice that I had to make Which was do I continue and do a phd or do I go and work in industry? And for some reason I decided that You know, I'd rather go work in the industry because I want to see how machine learning is used So I just want to point out that this is in 2006 So today if you say I want to work in machine learning, it's like, yeah, whatever Of course you should be working in machine learning Not so much in 2006. Also, I somehow decided that I want to work on neural networks I was convinced that there's a job out there That uses neural networks and solves a really important business problem. I didn't know what it was and I spent several months searching And eventually I did find a job in San Diego At this company called FICO or it was called as Fair Isaac back back then Now it's called as FICO. You're all probably familiar with FICO scores, you know, intimately familiar They had a division in San Diego that used neural networks to do credit card and debit card fraud detection And I you know ended up there working in this and But basically let me just point out that This particular team You know was one of the first ones to use neural networks to do this type of work So Somewhere in the middle between 2006 and 2009 Ashwini and I decided that we wanted to get married and she said I'll never come to the US So obviously I had to pack up and go back Though she's here now. So, you know never say never I think So I went back to India and I worked at ISE. This is what's on was mentioning. This was a Research group that was funded by this company called Satyam And it was inside ISE and I had the privilege of working with Professor Shalab Bhatnagar on Reinforcement learning and simulation. So this is what really piqued my interest to come back to USC And do a PhD with the Rahul Jain Professor Rahul Jain and Professor Amy Ward I thought I'd come back and work on reinforcement learning and simulation but I got really interested in game theory and queuing theory And part of my thesis was studying how people make choices on when to arrive at a queue And along the way I had to develop several queuing theoretic models Which is how I ended up working with Professor Amy Ward in the business school So what did I do to get this job? So besides, you know working on really interesting problems Thanks to Amy and Rahul I so this is sort of a pointer to the young people in the audience I reached out. So, you know around 2013 I was interning at Bell Labs And that was my the first thing that I started thinking about should I go into academia at this point because I was sort of convinced that I couldn't make it and You know on Rahul's encouragement and Amy's encouragement I reached out to several professors at Stanford and Berkeley Just to talk to them and give talks You know there and get their feedback on what I'm working on So I reached out to John Walden And Rhonda Ryder who are at Berkeley and Peter Glenn At Stanford and I owe Peter Glenn an immense You know debt of gratitude because he was the chair of the MS&E department at Stanford But he would carve out 30 minutes every Friday where I would go and talk to him about my thesis And I would you know, we would discuss problems that would come out of my thesis And I think before I went and spoke to Peter I wasn't convinced about the quality of what I'd done But after I spoke to him I sort of understand understood it so much better and I saw so much value in what I had done That when I came to give a talk at Purdue or wherever I think My talks are much more insightful and interesting and useful to people And I think that's part of the reason why I have this job. And so I I know oh Peter, you know an immense debt of gratitude Okay, so what about the road to tenure? I have A couple of decisions that I made so early on I had some success getting funded to study non-stationary queuing models This was all like sort of based on my thesis, but I decided to expand my interests in that area a little bit So I you know started working on a little bit of control reinforcement learning stochastic models But I also expanded the people that I was working with so I started working with folks in math and Business schools and of course in operations research So these are all some of the colleagues that I worked with I also received a couple of patents Just because I wanted to see how the patent system worked that was with Vijay there at the at the end And another decision that I made was to sort of Make sort of a decisive shift in my research interest So I decided that I wanted a new research direction in particular working on Getting closer to machine learning thinking about statistics and theoretical statistics a little bit And so I started I wanted to use like these probabilistic tools to analyze, you know machine learning and optimization So all of these things sort of lead me to like distill some knowledge that hopefully Is useful to people. I don't know if it's useful I thought I had come up with this but then I realized that Steve Jobs had sort of said something very similar in his commencement address so be hungry be foolish So be foolish take risks Like that's a decisive shift around 2017 the upside is that you learn a lot Work on new things all the time The downside is that there's a high probability of failure You you may find yourself on an island. You have to find a community And or you have to join a community and speak their language, which is a hard thing to do I found myself in that boat But you know, I can't begrudge any of the decisions that I made because it did lead to my career award As far as hunger goes I would say that this is the most interesting thing Try to work towards a purpose but enjoy the the process And this is a quote by Raghupashapati who is our colleague here in statistics whom I consider to be a faculty mentor And one last thing is ask for help, you know, don't be shy To reach out to people people are nicer than you think And I've been the recipient of kindness for many people And let me end there. Thank you Thank you for a great presentation harsha. So any question for dr. Honapa So in son's introduction, I think he mentioned some work on on a model to predict the success of students Yes, so what do you see as the possibilities for using these techniques to deal To improve the student experience in a program that's as large as ours. Yeah, are there opportunities there? I think there are yeah, so so the thing that i'm working on right now and we made some progress on it is You know when I think about ie for example, right? So we have I don't know how many undergrads is it 800 over 800 over 800 And I think we have four academic advisors, right? So just think about the You know the amount of pressure that they're under to provide, you know useful advice And the other problem that I see is students It is sort of a We need the students to reach out to us for advisement But can we sort of turn it around like can we sort of make it predictive so that you can reach out to students saying Hey, you know you need to improve this or maybe you're really good. Maybe you should consider going into grad school You know or maybe take these courses. It'll help you be more successful as a grad student so how do we do this in a more predictive way and One thing that I've noticed is uh students have a profile, right? I mean the certain types of students can be sort of Sustest out from data And so we've been building some machine learning models like credit sort of like credit risk scoring models to predict the success of failure of students this Give a probabilistic assessment of whether they are likely to be successful or you know, are they likely to fail out or You know, how do we and you know using this to sort of In a sort of predictive way so that we can reach out to students I'm in that way sort of you know reach that scale that you're talking about Yeah, and the other thing that he talked about was This codex stuff like so the gpt Well open ai has this thing called as gpt codex gpt3 codex So you enter natural language text and it produces Automatically produces code like you know computer code So we use that to solve all the homework problems in in optimization and production systems um And so you know one sort of friction that I see in our students ie undergraduates is You know, they're not necessarily the best programmers, right? Because they're there to be industrial engineers not necessarily to be computer programmers So how do we like help them with that? And you know if can we use codex in a way that it gives them it helps them sort of write better code And so they learn better, you know on the modeling front, which is what we want them to Pick up Any other questions? Yes Very exciting talk very inspirational I did notice you talk about around 2006 You tumble into data mining and neural networks And it's my recollection around that time neural networks were dying or dying And for a lot of students many of them sitting in this room Perhaps they're also dreaming of the next big technology wonder and so can you talk about What made you continue? Along the line of machine learning and neural networks. Yeah, um, that's a good question. So I mean as I said, I'm kind of foolish All right, so I tend to do these things at that time I Um, I was just convinced so there was a job that somebody was using neural network somewhere Even though it was sort of a ai winter. I think at that point I honestly don't have a good answer for you that I just I was just convinced that that job existed and I hunted You know and it was there. They were using neural networks to do it um I think you can't get too Caught up in current trends is what I would say like is something fundamentally useful is something You know like for example when I think about Like chat gpt right like today like this is what everyone is talking about I think chat gpt is kind of a fundamental tool To me it feels like, you know, I'm old enough to remember 1999 and You know experiencing google for the first time Because the people too young alta vista was horrible like so bad Like you would enter a search query and you would only find the answer to what you're searching for on the fifth page of the search results So google was like a revelation and I feel like chat gpt is sort of something like that Now should you work on chat gpt or you know large language models? Maybe right. I think there is an interesting, uh, you know problem That's immediately apparent with these types of technologies, which is You know it takes I read somewhere that it takes about seven times as much Computing power and energy To answer a chat gpt query compared to google search Right So there's there's an interesting engineering problem there. Like how do you like shrink that? So there are things that you can work on like even though this sounds like something that's You know too hot right now. I think there are fundamental questions there and I think these types of technologies are fundamentally useful Okay, I hear time's up. Okay. So thank you Dr. Honapa look forward to seeing your continuous Continue the success. Let's give a big hand to dr. Hush