 Okay, so good morning everyone. So it's my pleasure to welcome you to this celebration of our associate professors So for those of you who don't know me, I'm Mark Lundstrom. I'm the dean of engineering this year So it's a pleasure to be here as I reflect back on 40 years as a faculty member I think my time as associate professor was a very special time So it's an opportunity to learn how to define a vision or important and original research How to raise the funds to support the work how to mentor graduate students How to produce and publish important results and in the process you earn the respect of your colleagues and peers so it's a very special time and Now you have an opportunity to reflect on what you've accomplished and maybe accelerate along the same trajectory or Or maybe move what you've done into a new direction So it's a it's a special time in your career now We're going to have three presentations. We ask you to present. I think a ten minute talk We ask an awful lot in this ten minute talk We ask you to convey what you do in a way that is broadly accessible to experts outside of your own field maybe tell us a little bit about the Educational activities and engagement that might be connected to your research We're interested in the path you took to get to where you are today And maybe your vision for where you might be heading in the future So with that, I think we'll begin with our first speaker and I believe Abhi is going to do the introduction Well, thank you Well, I'm delighted to have the opportunity to introduce professor Waneet Agarwal Waneet joined us in 2015 after a five-year stint at the AT&T research labs Which he joined in 2010 after he finished his PhD at Princeton So he overlapped with Mung at Princeton. I believe a little bit So Waneet has been quite active since he has come to Purdue come to industrial engineering He also has a courtesy appointment with ECE He had played he also played a key role in starting the Intel Purdue cyber security designed for security badge Which fit nicely with his research area that he'll talk about but his overall vision overall research area is in Managing data and being able to make good decisions When you have data that is distributed across networks. So Thanks a lot. Oh, you already have. Yeah All right. Hello, everyone. Um, so Hello, everyone. Yeah. Hello everyone. So I'm going to talk a little bit about the research and other activities that I've been doing here So as we just introduced Things about background. So I graduated in 2005 with undergrad from IT Kanpur where I received the best student awards. I Did my PhD in 2010 from Princeton with electrical engineering majors CS minors and affiliation with math And I got some awards there After that I spent four and a half years at AT&T research labs where I was working on topics in networking and machine learning and Since January 2015. I've been here and I'm still waiting for getting some awards here And I have other adjunct appointments at bunch of places which I also mentioned here Okay, so a brief about the Google Scholar profile per se of how the track has been over the years 2015 Jan is kind of when I joined Purdue and Thanks to Purdue and my past experience at AT&T things have been going well so far So I have about 87 journal papers 97 conference papers accepted 20 US patents and one book chapter I have received the journal best paper award in 2017 And a conference best paper award in 2018 all of them after joining Purdue In addition, I have presentation awards and I'm writing a research monograph Which should be out Likely end of the semester Okay, so in terms of research areas. We are mainly working with aspects of big data Now big data involves multiple aspects One of them as you all know is the aspects of machine learning where you have lot of data around us You need to make some autonomous decisions. You need to mine data or things of that But other aspect is your data will be stored on the servers So with the distributed data comes lot more challenges of how to store how to access how to process So that is in summary the type of things I work on so and How to store how to access those topics are traditionally like networking aspects And how to process and how to make decisions come under AI So in that is essentially why in summary the types of topic are Combination of machine learning networking and cloud computing, but even in networking cloud computing It's basically how to act how to process and access the big data Okay, these are some of the Detail topics which I'm not going to go over I'm also working on autonomous vehicles and smart grid type aspects which are also related in a way to big data because vehicles autonomous vehicles have are Expected to increase as the days go by just to go a big detail about some of the projects So I'm using this quad chart which I kind of used a few years ago to explain the research in the EC primary committee, I guess So the first topic that we'll go in detail is related to cloud storage and Abhi might know this was the topic I presented during my job talk here So the cloud storage essentially is All of us have used Dropbox Google Drive box things like that We put our data on the cloud But then when we access it we want to get it accessed fast But there are many challenges with that When you're storing data up on the servers you need some redundancy Because the disks fail all the time Now if you have redundancy that gives you new opportunities of where to access the data from and That problem is very hard All the file systems today like quick file system Tahoe Hadoop Facebook and all the companies Google file system All of them use some redundancy But we need some efficient techniques of getting the data from there faster the better Okay, so we have been working in this type of space to come up with efficient algorithms to get the data from the servers as fast as we can some of the sorry Impacts that we have achieved we have demonstrated that we can reduce the latency latency is the amount of time you take to get the data By a multiple of five as compared to current implementations on the current file systems in addition to this improvement of latency just playing with the optimization we can do efficient caching strategies caching is When you have a Dropbox or Google Drive folder you put some of the things on your local host If you can manage that local host well We can reduce the latency further by 25% as compared to existing things on self caching module So we beat most of the state of the art caching systems by doing a fundamental analysis of the storage systems and the overall vision is We need to come up with the overall system for Distributed file storage systems that care about latency Reliability and the price of access and in order to care about all of these aspects You need to understand the system end-to-end local parts of the systems are not sufficient So that has been the vision and this figure is kind of seeing you the different Tuning knobs that you have that you need to consider all of them together The second project I'm going to talk briefly about is on ride shearing Uber and Lyft have been increasing over the years there been lot of demand of them and that is going to increase the personal vehicles Will are expected to reduce as the autonomy kicks in more and more With that perspective we need efficient algorithms for matching the passengers and the vehicles Okay, so even though Uber and Lyft are doing their jobs their algorithms are not known and Who knows whether they're doing an effort a perfect job So our aim in research is to come up with more efficient algorithms One of the thing that I'm going to explain is we came up with a new methodology like in uber pool They take a passenger from the source to the destination what if You have two riders Rider one is taken by vehicle one to here and then this vehicle was going to be and then pick up the vehicle The rider one two so what is happening is the passenger can get dropped in the middle to be picked by another vehicle That happens commonly when we take the buses and trains Why can't we have that in the ride sharing systems? We implemented those type of strategies using machine learning and we showed that Contrary to what people think about the negatives of that. This was very much helpful The other type of things we are working is the concept of reinforcement learning which is autonomous decision making The concept of reinforcement learning is you have a state You take some action and then you get some reward and how to manage with that is another thing that we are working on Some of the other things with machine learning that we are working is you are given a missing data like the Netflix rankings or Where people do not give you the rating of certain movies. How do you fill in those data? Without going over detail, I'll just show you a small video So we took this video of the YouTube where a gun is being fired You see the hand gun smoke and the bullet We removed 90% of the pixels here Okay, so you get almost nothing if you remove 90% of the pixels in the data and Our algorithm can reconstruct this thing with 90% of the missing data So essentially you don't even need 10% of the data in most of the images and our algorithms could have that much impact Which you can see is the efficient way of data compression in a way Okay, the research here could not be complete without all the faculty collaborators that were that I had at Purdue With a combination of IE and ECE faculty and thanks to the Purdue EFC programs that we could also go to other collaborations with other engineering programs and also other colleges Beyond engineering and special thanks to all the other faculty mentors Who have been helping me to go through the procedure in terms of teaching? I have mainly been teaching the courses of stochastic processes Undergrad and grad as well as machine learning I've also been teaching courses on scheduling in computer systems which includes Combination of approximation algorithms stochastic processes and machine learning and combination of them for efficient scheduling and planning So that's all I have thanks to Special thanks to all the students that I have worked with at Purdue and elsewhere So we have graduated four PhD students from the group There are seven PhD students at this point three of them are co-advised With some of the faculty in the room I've had postdocs Mented master research and have multiple undergrads in the group Thanks to all the Purdue programs including summer stay surf the VIP program Which which is started in the ECE and has been a great program sorry Purdue IE senior design project course and Some other programs on that sign and then there is a Purdue horizons program I don't know if you know about that. That's for first generation College students as well as minority. So I've also been Faculty mentor for that and have supervised some students through that program. So that's mainly about The student side of mentoring side of things That's all I have and I would look forward to your questions You know patenting your algorithms, how is that whole interaction worked out because I can imagine there could be a lot of interest Yeah, as I just mentioned I have 20 us patents. I think 14 of them are granted at this point The issue with as such algorithms is that algorithms are harder to patent So we do not get like the data completion algorithm We cannot patent it but because it's a standard type of means it's an alternate minimization type algorithms So some of the algorithms are harder to patent while when you talk about this Distributed storage setup where you basically say these are the flexible knobs We can expose and set up a problem like that that technology with the algorithm can be patented So that is essentially how we have been approaching So we have mainly the patents around networking side of things Not really on the machine learning side of things. I don't have a question, but I have a comment When it has been I'm I'm I'm Bharat Bhargava in computer science department and I've been here For a very long time including in electrical engineering when it has been very helpful to More than four or five of my PhD students and I invite him to serve on committee And he was very instrumental in getting a multimillion dollar DARPA grant he's So unselfish that he's willing to help me with any projects that I have particularly the new one that we started on with Ford Motor Company, so I think If you're a student you want to get involved with him If your faculty you want to get involved with him and if you're the dean you want to promote him Thanks a lot any other question. All right. Thank you