NVIDIA Senior Applied Research Scientist
UC Irvine Ph.D., Information and Computer Science
Current Time 0:00
/
Duration Time -:-
Progress: NaN%

How did you get to where you are today? What is your story? What incidents and experiences shaped your career path?

Summarized By: Jeff Musk on Mon Feb 17 2020
Okay, so my story ah is basically I, uh I am originally from Ukraine, so I got my bachelors in applied mathematics and the Masters in a bad Miss magics in the ah, you grand University of Give. And, uh, over there my major was a brightness magics, and I was doing, like, graph theory. So then I applied to the U S and got it into the PhD program at the University of California. Your line. Ah, where I said the chance basically to work on applying graph Syrian statistics to the problems in buying for magics. And once I was ah, doing my PhD coursework and research, I kind of got really interested in machine ronin eso. After I finished my peers d I got the job, but Microsoft, as in ah, applied researcher. And then after that, I spent maybe you extend 1/2 years Microsoft working on some projects inside, being search engine, and then the ice, Which toe EPO where I worked on like else, isn't applied researcher on the condition programs for up store in the items and the basically no limit on video. And I've been here for, uh, three and 1/2 years. Andi, regarding experiences, I think important, obviously, educational experience is very important. But I also would say, like because my original education wasn't more for pure a pride must. Ah, But during my student days, attorney Joseph Keefe, I got the chance to participate it. Ah, programming competitions organized by Microsoft Gold. Imagine Cap. So I guess those experiences kind off convinced me. Then I kind of moved from masked, more off a computer science bus.

What are the responsibilities and decisions that you handle at work? Discuss weekly hours you spend in the office, for work travel, and working from home.

Based on experience at: Senior Applied Research Scientist, NVIDIA
Summarized By: Jeff Musk on Mon Feb 17 2020
Ah, yes, sure. Uh, certain video. I'm ah, produce your scientist and the I am part off actually off a product group, eh? So you're not part of a media research. Ah, but we also do, ah, bright sense in the sense that basically we work on their problems The tolerated toe, what we call conversationally. I So this involves especially cognition, natural language processing. Ah, speech generation. And, uh, we write a lot off open source quote, which we share with our partner. It's open source special shared with everybody and the ah, the go off. Like why we just given a very cold for free is basically toe grow GPU usage for deep union. I make it easier for everybody to use. So basically, since him measure go off our work and, uh ah, we also do research in the sense that we develop new models that I run faster or maybe useless memory or maybe achieve better accuracy is so when the good something interesting like that, we open sort we publish modern, right? Go to the conference Insulin. Ah, in terms off weekly hours. Right. So I would say my my work and this is why I'm Kodak Applied scientists is a my work typically cycles between development and doing research. So once we have, once you have signed my year, you basically need to implement it, right? So And typically involves a lot off engineering, especially for if your idea involves something like, uh, um, drinking onwards off nodes, works of GPU. So there is some tough to engineer an aspect you need to handle first. So once you're done with that this and walks your work experimentation stage right where you can run experiments, dr various like no union rates, model considerations and so on, Basically kind of cycle between those toe Moz, Uh, in terms of ours, uh, I think it's true for no turning foreign video, but for money. Ah, companies in Silicon Valley that, like where the time you are actually like we don't have Ah, we don't have any hard requirements that they must come in it like nine livid whatever five or six. So we don't have anything hard eggs that, uh, I mean, typically, most meetings happen, like in the middle of the day, but it's totally fine to work from home when you have to, and you can come wait and then leave right or vice versa Come early, leave early. So it's it's pretty plan for civil authority.

What tools (software programs, frameworks, models, algorithms, languages) do you use at work? Do you prefer certain tools more than the others? Why?

Based on experience at: Senior Applied Research Scientist, NVIDIA
Summarized By: Jeff Musk on Mon Feb 17 2020
Okay, So in my current role, uh, in terms off languages, so we mostly use bison because this is, like the most report is the most popular language for deployment right now. Ah, So and you need to, like typically with four different, you would use some framework. Ah, in ah, my day to day work. Right now we're using pytorch. Ah, we were using tender four before, Um, but I would also say that this is very flexible. It's actually I mean, it depends on a particular researcher. What? Hugh, she wants toe use standard four by torch or a max in it. Or she indicated Whatever. Uh, actually, a few years ago, I can defend this policy for myself that I just try new project with new framework. Um, yeah, but now it's mostly pytorch. We also sometimes, especially if you want to do something like, ah, and there's a code of the framework. Then you are actually out of the bison territory and into off c++ Cuda territory. So, yes, I know what you're for. Ah, more whatsoever. When which, such as the super sports, which she really hopes. Ah, yeah. So

What things do you like about your job? Were there any pleasant surprises?

Based on experience at: Senior Applied Research Scientist, NVIDIA
Summarized By: Jeff Musk on Mon Feb 17 2020
Yes. So there are actually so first of all, I mean, I I like the work. I do So the kind off work problems your work on, but in terms off pleasant surprises ex specifically to any media. Uh, I would cry rightto 1st 1 is it? It's very like because I felt an experience in other companies. Ah, at the video, it's like it's surprisingly easy for me to, like share in the open source, our work. So that's, like very easy efforts. Very easy to go. You, like conference and discuss and shares. Yes. Ah, it's a I guess it's influenced by our business model because we sell hardware and then the software and research. Yes, we kind off mostly give away for free so that it would drive, Ah, demand for the hardware, huh? But in my particular role, it makes it very easy. It's basically a Ziff. I am in a Virginia I can share, like actually the project you're working on right now. We just even develop in the open on, like NVIDIA get trapped. So it's open source. And, uh, yeah, it's very easy to work with some other companies on a particular problem. So that's the first surprise. Uh, another one is probably not a very big surprise, but very presence in is naturally, uh, this being in video, if ever Ah, access to a lot of cool hardware used for deeper union. So that's also very important nowadays for different research.

What are the job titles of people you routinely work with inside and outside of your organization? What approaches do you find to be effective in working with them?

Based on experience at: Senior Applied Research Scientist, NVIDIA
Summarized By: Jeff Musk on Mon Feb 17 2020
uptight. Those, I mean, is a very oh, awards, right? It depends on the company. Ah, uh, to give you some examples, I would say I mean, if if we're if we're at some come, she ate some conference, and we want to, uh, talk with the company. If this is a relatively small company, there may be like a city. Oh, even see you in the meeting. Ah, but if it's a bigger company or this is a room, you'll be, uh, as the folks, I would say, more important thing is not a job title, but like a like a job role. So I would kind of distinguish between engineer American business. Right. So So inside the inside engineer in doesn't really matter. If it's like junior engineer, senior engineer, distinguished engineer, and so on. They kind of speaks the same language. Um, but if you have, say, business people and, uh, you represented genuine people, then you wouldn't need toe. Maybe a kind off, cried some low ever details, right? And try to come up with a however off kind of representation of what the product has to do. So you kind of have to switch your head a little bit from Mike Engineering Toe product so that you speak the same language. So I would say that the room is more important when you talk to peoples and their specific title.

What major challenges do you face in your job and how do you handle them? Can you discuss a few accomplishments?

Based on experience at: Senior Applied Research Scientist, NVIDIA
Summarized By: Jeff Musk on Mon Feb 17 2020
uh, major changes. So we work in the deep Iranian feud. So this is a field that is moving really, really fast. So it's both exciting. Ah ah, sin. And, uh, sometimes it can be challenging. Ah, just do like a specific example for, like, an accomplishment that so we were working on the like Last year 1019 we were working on a new model for special commission and, uh, with developed its train it and it took pretty good, like on the benchmark you were, ah, mystery were between the benchmark getting the best result. Ah, with admitted to the conference. Submit our paper to the conference and there is a sight like where you track the progress. Which organisms is the best on particular benchmark. So we were number one for two weeks. What's happened is another team at Google. Actually, they submitted to the same conference like bitch in ours out a little bit. So, uh, both papers got into the conference and it was great conference, but the series the challenge that the field is moving really fast, it's ah, it's ah, it's sometimes car took a chap and I would say that important part is not to try to catch up, but just to try to do your best work. You have some idea, like evaluated show if it's working or not, and try to understand why and that the 1 may result in like Valuable is a product. We're a research publication, even if it doesn't that you like strictly the best state of their.

What are the recent developments in the field? How significant are these improvements over past work? What are their implications for future research & industry applications, if any?

Based on experience at: Senior Applied Research Scientist, NVIDIA
Summarized By: Jeff Musk on Mon Feb 17 2020
So in in the field off conversation of Ryan Generals Reeses A very rapid progress in, ah, natural language processing. It's in the sense it's very similar to 13 15 few years ago in the image processing, basically when you can. But it rained some models on the big later says, and then fine tune them on different tasks on the four different data sets. So now this is happening in energy so we can train those future language models. Uh, actually is a very nice thing about those orange models. Is that their trend on unsupervised data? So there is like, Ah, turns off unsurprised text data out there owns the Internet. Wikipedia. So what's of data? And once we trained those big, huge models they get, we can reuse them. Waiter for the tasks that are, uh, the do require supervised data. But we have much wrestle that supervised it to such a state. For example, question answering and the is the sense you are training. The model not only owns us your questions and the answers, but on that whole what Wikipedia in knowledge. And it really improves those downstream tasks and works very, very well transfers toe another the mains and so on. So those Ah, very good. Very exciting advancements again over the state. So starts and baselines are changing very fast. And, uh Ah, in terms of their implications for the future of research and industry, I would say that we would see my smarter Ah, personal systems. Right. So seems like a let's ah ah, Google assistant cities of you will become much smarter and, uh, in terms of research, uh, we would be just we would be able to use viewers will be able to do so. Seems like, I don't know, maybe die work like more. Ah, like more dialogues that are the resembles people Look closer. So, yeah, that's a very exciting time for an O. P right now.

What qualities does your team look for while hiring? What kind of questions does your team typically ask from candidates?

Based on experience at: Senior Applied Research Scientist, NVIDIA
Summarized By: Jeff Musk on Mon Feb 17 2020
rehire, uh, for applied deployment Rose. So what that really means is that the candidate should have ah, two kinds of skills. One kind is a candidate fest to be actually a good engineer. Ah, like a programmer. So that means that strong computer science from dementia ls and, ah, knowledge off at least one object oriented language And on those offend because it's, ah applied deployment roll. The candidate should understand basic mass behind the deployment. And, uh, and some day brilliant basics. So those are two categories off porridge is that we are looking for. Uh, of course, we also Oka whether it's a they could ah, persons that is easy to work with, whether he or she is ah, employer. Uh, in terms of questions, Uh, yeah, I can actually tell what exactly the process is. So for internship positions ready quickly have to phone screens toe to technical phone screens. Before that, you may have, ah, just a phone screen with the recruiter. But when that phone skins are no technical questions, but on the two technical phone screens you can get program. You will get the programming assignment and the some Dobrynin questions. And in terms of programming assignments. We're not We never give people like a treaty like, Oh, Basel questions. Uh, those are the questions we give them. Is basically trying to test whether is a understand data structures was they understand Time complexity. Whether they understand where it makes sense, Stop. Why, for example, Set versus least versus Christabel. So nothing really tricky but trying to understand to measure their understanding off computer science from the mental sze uh, on the jeep alone inside. Ah, basically any since that Ah, the question we ask them typically covered by some, uh, course there, of course, that you can take on wine. So since about like, what is convolution in your own network? What is that? And then what kind off angry since is there to train them? So So this is for interns. It's to phone screen threat. For the full time, candidates would typically have who was three phone screens and then before as a candidate inside, where we actually have a full day off, like five interviews and several people needs him. So because that would be like a full day off intelligence

What was the hiring process like for your job? What were the roles of people who interviewed you? What kind of questions were asked?

Based on experience at: Senior Applied Research Scientist, NVIDIA
Summarized By: Jeff Musk on Mon Feb 17 2020
process was exactly like conspicuous question. So basically had the we started with the phone's screen with Recruiter. And then there was There were two phone screens with the engineer. Sends a team. Ah, during during those phone screens. Basically, I was asked some basic questions about what I was working at that point and the some questions related to a capitalization algorithms. After that, I said the one full day off on site interviews. Uh, I think I met with five people. Ah oh, and it's Ah, it's also important North. And when you go for their own side for the full time role, you know you always meet with your credit manager. Best visit person who you'll be your manager if you if you have. So I met with my perspective manager uh, and the I met also with what? Like three engineers who are now my co workers. And basically, I was asked. Casey pushed class programming question one person programming question in another interview and the like several deep alone and systems designed questions. Uh, actually, my injury was a through process was a little bit different because ah, at that time, both the team I was interviewing four and me. We went to the same conference, so we actually felt like one and off interview during the conference. And this is this actually happens frequently for if you go to a missionary in conferences.

What are different entry-level jobs and subsequent job pathways that can lead students to a position such as yours?

Based on experience at: Senior Applied Research Scientist, NVIDIA
Summarized By: Jeff Musk on Mon Feb 17 2020
so you would start with its very, very helpful to start with an internship. Uh, actually, on our team we have right now with me. She at least at least three full time people who actually started as interns. Ah, so one people was an intern on Not on our team, but Steven kind of related internship in media. So this is Ah, very helpful past when the student, during their strategies, can come and stay for a summer doing internship and then once they graduate Ah, we can, uh, come back for food. 10 road. Uh, it doesn't have to be Well, exit. Uh, you can if you are a new college graduate. Ah, a new media typically goes toe several campuses. Ah, to hire new college graduates so you can apply on your career fair if you see any year on your campus. Uh, if you don't see in video new campus, it sze Totally fine toe. Apply yourself. So, uh, you can basically get hard out off any school and like even any country, so that doesn't really matter, huh? In terms off subsequent job, past ways I stayed the during my career I was I I was always into engineering. So for me it was only whatsoever. Wasn't region like? No, the social one researcher toe senior researcher. So I Sometimes I know people switch the disciplines like you can go from engineering to, say, program management or resting or from testing to engineering. So that happens, too. But for myself, I just always stayed in like this. Applied research past.

What were the responsibilities and decisions that you handled at work? What major challenges did you face in your job?

Based on experience at: Principal Applied Researcher, Apple
Summarized By: Jeff Musk on Mon Feb 17 2020
a tempo. I was working on recommendation algorithms for up store and the iTunes eso for basic Moses means sometimes when you go on the APP store and then say you say, for example, you click on Expedia. Then there is a tab like it's called Related Assumption exists where it shows you. Oh, customers. Elsa bought Travelocity or American Airlines up so basically uses. It is psychological relation between x pds and say American everyone's So it was one that was probably was controversial. Future. Another feature I have worked on his was, ah, game recommendations. Basically, it's when you go to the APP store and becomes the game stop. It would show you something like more games you might like. So this is basically recommendations off games personally for you, based on what you have already purchased. Ah, and so this is a personal is the condition feature? Ah, various customers off support. It's not personalized because it only depends on the context. Uh, so my responsibility as an applied researcher to develop algorithm behind this future and I also helped this engineered make implementing some, like drunken features implementing some of the data pipelines necessary toe make this work. Uh, so what was major challenge? So it's a major challenge Was actually was personalized recommendations. So as you probably 53rd EPO is very, uh and this is a very good sense. So EPO is very careful about user privacy. So we were via when we developed this feature we were able to use only basically only absolute you already purchased. We didn't even know what you say. Uh, uninstalled what you using? How often? Because that involves, like, privacy. We like user's privacy. Ah, so that was one challenge and as a big challenge was actually because when I started, that was a new feature. So that was the first personalized recommendation featuring the APP store. So it was very hard toe convince a jitters Capstar editors and business people that this is something that we could is this official ship to the customers, that it would be a good user experience. Ah, nobody really adopted that it would increase revenue. But the very important part was like convincing a lot of people that the user experience, you'll be good and the lady convinced off them. And then it was basically we did what we call Doc. Fujian is basically the ship, the product, but only inside our team. So one inside Aleppo inside the team. So show them and directed. And then we did the bigger test, and then the bigger test. And then we started with, like, a B test, which means we were testing on parts off, like US traffic and then the ship to the whole U. S. And the new case. And sometimes it was very gradual process.

What were the responsibilities and decisions that you handled at work? What major challenges did you face in your job?

Based on experience at: Research Software Development Engineer II, Microsoft
Summarized By: Jeff Musk on Mon Feb 17 2020
Chris off the I was working on several since one very cool scene I worked on was actually, it was 22,013 and we worked with with another person from Microsoft Research when ah, on the deployment took it, actually. So it was my response. So they called it a shanty. Kay and ah, it was my responsibility to implement Ah, first GPU back and look for it. So, uh, yes, so basically is meaning. I felt toe duos those It's like magics operations and the congressional operations on GPU implementations and the throw the major challenges where that it was Actually, at that times it does. Ah ah, very well. Wherever for this kind of tasks, for example, could you And then what if I breathe to ship it in video right now? Didn't exist at this point. So we had to do a lot of those front functionality, just myself and the s. So I had to help with this. This was one part of my job and another part off my job was to kind of a pry our usual different and took it on the rial problem at Microsoft. Ah, which at that point I was doing was, uh, really great predictions for being cards. Uh, and there's a big challenge. Was, is that the basement was actually relatively pretty, pretty high already because what off? Very smart people worked on it, and it was very hard to build that to be that baseline.

How did the program prepare you for your career? Think about faculty, resources, alumni, exposure & networking. What were the best parts?

Based on experience at: Ph.D., Information and Computer Science, UC Irvine
Summarized By: Jeff Musk on Mon Feb 17 2020
PhD student, right? So very important. Part off. Here's the experience is ah. Ah, like your research and the what kind of advisor do you have? Uh, even though my research, my PhD research is not directly related to with what I do right now, is that the process of doing the research? Yeah. Like learning howto define the problem. Outlined. Promising ideas like rank them, which are promising region north in terms off benefits versus home hard would be to try Samson. So this is this, like, research process in general. So learning it was pretty useful because now is an applied scientist. Ah, I can't just, like, sit and think about something for half half a year without producing any results, right? I need toe identify most important directions. Split them into smaller chunks that divisive, planned, how to proceed. And what what would they quite do if something doesn't work as I hoped it would, and so on. So doing appears you really prepared me for that. Another, like, earlier part off part of PhD where you take courses like algorithms, courses, probability statistics. So those are off course. Very useful. But you don't have to go to PhD program toe. Get those courses. Ah, yeah. Explosion, Networking. I really laughed. And they still have goingto conferences doing like research presentations, networking with people. So those are basically where's the best parts off PhD program? And, ah, they're one of the best parts off my job right now. Just go somewhere, Ah, several times a year to meet with different people, share ideas and do some sort of presentations off our work. That's that's very nice.

Would you like to share something that is not on your resume? This may include your passions, facing setbacks or adversities, a unique experience, or an unexpected help.

Summarized By: Jeff Musk on Mon Feb 17 2020
Really? Yeah, maybe. And that I'm expected so bright. So and this is actually related toe. Ah, my career past. But it's not my business. So when I was at the high school, uh, it's just by pure chance I met Ah, like a very great mass teacher. So that was, like super important for me being able to even gettinto Kiev University and so on. So she basically, uh I mean, if you left his job so he helped us tow prepare for different like mathematical contests and so that it was super helpful. Ah, and kind of a unique experience because, I mean, they didn't meet him then. I would probably like, not have as many like most magical contests and heard the West chances of getting into the Kiev University in Seoul. So that was, like, a super helpful to get this sort of help. Ah, very early on my past. Yeah. I mean, in terms of passions outside off work, I can I mean, I just like a pretty much have resonated toe outdoors. Ocean. Uh, yeah, it's important to not forget to have fun. Uh, so, actually, terms of passions once and I would add this, uh, when I was a piece of the students. So Ah, and this is one of the very good parts of being a PhD student. So you manage your time yourself. So and I actually spent a lot of time on the beach surfing as a PhD student. So my advisor was totally fine with this. As long as I produced eight years and results and the it kind of helps like being outside kind of helps you come at least for me. It really helps toe come up with new ideas off course. Sometimes you have to, like, be methodical and the just sit and work for extended periods of time to get something done. But another times, you just want to refresh your mind and get new ideas and like being outside really helps.

Do you have any parting advice for students and professionals hoping to get to a position such as yours? What 3 dos and 3 don'ts would you suggest?

Based on experience at: Senior Applied Research Scientist, NVIDIA
Summarized By: Jeff Musk on Mon Feb 17 2020
So if you are a student, wantedto start working in a similar position. Ah, like I said, Good, good First step would be trying to do an internship and the How can you get an internship? Uh, let me give three. Does right? So one is It's very helpful toe for you to have some sort off open source presence. So if you have a get have account where you share your open source project projects or maybe you contribute to some open other open source projects, so we go look at those. So it's it's very important another person is. Maybe you are same as my student or engineering students who wants to start it like kind of, ah, check out deplaning. But it's not directly your major. So one important thing that you can do is to take some course era or udacity different specialization, because I would say for me and for what off my cold looks. It actually doesn't matter whether you took a dip alone in course at course era or it's Stanford or it seem you right. So once you go to exist, age where you intervene with us, what matters for us is that actually did take the course and have the knowledge. And actually, when I see that the people took someone one course, it shows me that the person is like, self motivated. So it's a very good sign. Ah, yes, sir. And no, sir do would be to stand out from the crowd, basically try toe, participate in tow in some sort off, like maybe programming competitions or data science competitions. I mentioned that myself I participated in Imagine cup, so that was certainly helpful. I mean, after that, I was immediately interviewed by Google and Microsoft, but it was It was also very great experience. So it's very useful. Ah, three don'ts. Eso what? Missing about don'ts? Well, I mean, certainly don't Why? Oh, don't exaggerate. Since on your regime ear, because that's very easy toe catch. Ah, what else? Uh, three don'ts. So yeah, I guess that's the biggest Don't, uh, just, uh oh, let me get another. Do so actually make sure that when you're interim interview is a two way street, right? So you you're interviewing for the position. But you also should be kind off interviewing you. Enter yours in the sense that do you really want to work with them like try to understand the what's the role would be like wood. Would you enjoy It? Wouldn't be useful for your future career, so you can try to make it try to make interview as a two way street.