Merging AI Innovation and Practical Application with Dr. Vinesh Sukumar

Sheikh Shuvo:
Hi, everyone. I'm Sheikh, and welcome to another episode of Humans of AI, where we meet the amazing people building the tech that's changing our world today. We have a very special guest, Dr. Vinesh Sukumar, who's the head of AI product management over at Qualcomm. Thanks for joining us.

Vinesh Sukumar:
Thank you, Sheikh, for having me on your podcast.

Sheikh Shuvo: Yeah, absolutely. Now, Vinesh, one of the earliest questions I have for you is that you've obviously had a long career across different aspects of AI and done some pretty amazing things, which we'll definitely dive into. But if you had to describe your job to a five-year-old, what would you say?

Vinesh Sukumar:
I would say, you know, I'm not solving world hunger, but trying to make an attempt in that direction has got plenty of problems, solve one problem at a time.

Sheikh Shuvo:
Well, you're, uh, with the things that Qualcomm is up to, it'll certainly make a dent at least. Absolutely. That's the goal. Yeah. Well, tell us about your career overall and how you landed where you are. How did you make the decision to focus on AI hardware?

Vinesh Sukumar:
Yeah, sure. So, I started my professional career, give or take, at Dead Proportional Labs, where one of my first jobs at that point in time was to classify images. It was large images taken from a camera, and I had to classify trees, objects, buildings, gardens, those kinds of stuff. Those days were, I would say, our early integration into computer vision. And also into classifying things. Now the issue was, you know, it took me about three days to just get it completed using a magnifying glass, a ruler, and a pencil classifying stuff.

I had some information about geography and biology, so it was easy for me to classify stuff. So I go to my manager and say, you know, hey man, this is taking a long time, you know, and I simplify the stuff. He says, no, we are paying you. You figure it out. Okay. So kind of entered into, you know, assembly language. See, I had some good experiences there.

Started writing some code and trying to understand how can I use simple code to start looking at classifying objects? That was my, I would say, my entry into automating computer vision. Those days, visually looking at visual things and slowly with time, migrated into consumer products. Worked at Lenovo for some time in China, and then migrated to Israel, trying to work in an automotive environment and PC environment.

So kind of get a perspective of different consumer devices. And then my interest, you know, grew into a passion, really trying to understand what is the end goal? How do you make user experience much better? How do you automate stuff? That's how I started getting more into AI, even though I had my Advanced degree in computer vision.

Those days, it was very theoretical knowledge, right? It's not really practical, not put into consumer devices. That's how I started working on it. Grew interesting. This amazing opportunity came up in Qualcomm where I was given the privilege, I should say, to think big. And then obviously working with some fantastic colleagues of mine.

You know, translate that blue-sky vision into reality, right? So that's how I do AI pretty much these days, AI hardware, AI software, AI systems, all of those things.

Sheikh Shuvo:
Can you share any public-facing things that your team at Qualcomm is working on now that we might be seeing in the next couple of weeks?

Vinesh Sukumar:
Yeah, sure. I mean, these days there's been a lot more interest in generative AI. Even my 65-year-old mom seems to be excited because it somehow removes the wrinkles in her face whenever she takes selfies. So that's exciting times. So, generative AI seems to be making AI relevant to all age groups. Quite recently, I would say, about six months ago, at the Mobile World Congress in February of this year, we announced to the world that we have stable diffusion working on the edge on a phone-based form factor. Historically, people thought, you know, if you want to do any kind of text-based image generation, it was only based on DALL-E. It was only working in the cloud using OpenAI.

But we showcased that, you know, you could use open-source models, condense it, running it on the edge, and we were able to get it done in about 10 seconds. The same thing is, you know, trying to do image-to-image transformation. For example, you capture an image, you want to replace the background based on a textual command, based on a voice command, you can do it instantaneously.

So, you know, you're going to see commercial deployment of some of these most common applications going out next year, uh, in consumer devices like smartphones. And obviously, it'll scale to other form factors like PC, automotive, and those kinds of, you know, verticals as well.

Sheikh Shuvo:
That's all exciting news.

It generally seems that the pace of innovation in the AI space is nonstop. Right now, there are 200 new releases every single day from all the big companies. How do you stay up to date with the latest developments and figure out what to apply to your work at Qualcomm?

Vinesh Sukumar:
I think, from my personal experience, what is relevant today is obsolete tomorrow in the space of AI. So I think the most important thing is for any person working in the space to be a constant learner, a constant student. And if you're a constant student, you would always look for resources where you can educate yourself. And this could be attending, you know, pretty good technical conferences.

There are quite a few of them depending upon your technical acumen, which conferences you want to attend. There's a lot of articles that get published on YouTube and on LinkedIn. If you subscribe to the right newsletters, try to understand what problem statements are. And I know there are quite a set of interesting meetups that happen at least in the Bay Area where I'm from.

Then people start discussing about, you know, issues, practical applications, how they have solved it, you know, and getting different perspectives. So I think as you start reading this constantly, being in touch, you try to understand, okay, people have done things differently. And then we could use that exposure and experience to self-educate and then possibly apply that in your own application development as well.

Sheikh Shuvo:
Yeah. It sounds like you have a very packed schedule overall, but do you still have any time to tinker with AI projects on the side as an individual?

Vinesh Sukumar:
Oh, absolutely. You know, the most important thing is, I have two growing daughters, but my expectation is to be part of their life as much as possible and get them excited and keep them away from the idiot box as much as possible, in this case, AKA TV or a laptop.

So, I try to give them some interesting projects, obviously based on my exposure to programming languages and to AI. Can they use that to do simple things? And sometimes it flops, that's fine. You know, at least you learn to crawl first before you learn to walk and run, but it at least lets them know what is possible using AI.

It could be as simple as, you know, getting into robotics using AI as an example, or a simple artistic drawing using code language. So, as long as it keeps them occupied and finding it interesting, they get it going. So that's what I do. I'm pretty sure things will change as my daughters grow up, but at least that's my approach to how to practically use this in their life.

Sheikh Shuvo:
Yeah, no, that resonates quite a bit. I have two young girls as well, they're six and eight. So, how to introduce them to the world is something I think about all the time. Looking more broadly then within the team that you manage focused on product management, obviously lots of different types of people are needed to make that happen.

Outside of technical proficiency, what are some of the skills you think the best people on your team have that allow them to differentiate themselves and keep growing in their AI careers?

Vinesh Sukumar:
I think a few important things, right? Obviously, the technical acumen is important. You need to be a student at heart. That's critical to be successful in that space. But I think the second thing is definitely having the right personality and character set to be successful in the space. One thing that actually clicked with me personally was, I would say, always have a certain amount of dynamic tension with your engineering counterparts. It does not mean that you question everything, you fight with them for everything, but, you know, ask the assumptions, why these assumptions are good, and make sure at the same time that you have a lot of strong assets and background to justify a certain ask.

And if you happen to have a pretty good dynamic tension, I'm pretty sure the engineering community respects that and loves it because they're getting challenged on why a certain direction has been taken, why a certain execution time has been taken, right? And if you're able to explain, hey, why certain product perspective decisions have to be made, why compromises have to be made and able to articulate that well, people start to respect you.

And I think that's important. The second thing is always be a good listener. You know, I try to give an opportunity for the people to give their state of thoughts, what their opinion is, and try to understand their perspectives. Right. Even before you jump in and say, you know, it's completely wrong, I think respect that.

And if you're able to get to that level of intelligence, I think, you know, it's going to develop some good handshakes and you kind of build on it. So I think some of these basic things have helped me. I'm pretty sure I could do more, but at least some basic tips.

Sheikh Shuvo:
Focusing more on those communication skills that you mentioned there during the interview and recruiting process, are there any particular types of questions or scenarios you like to take candidates through to assess those skills?

Vinesh Sukumar:
Yeah, I mean, I try to really understand their analytical thinking process. How do they approach a problem statement? But, you know, the question is not just about jumping straight into the right answer, but the thought process. How are they looking at certain things? How are they coming to a conclusion? The conclusions can be totally wrong, but that's fine. At least the thought process defines their mental acumen, I would say, to a large extent.

So that at least gives me a good foundation. A second thing is, I would probably look at, um, their cross-functional team skills. Because to be successful in the space of AI, you've got to work with a variety of people, different personalities, different characters. How are you able to gel with them?

And have they had any interpersonal challenges? And if they have, how have they overcome that? Having some practical examples would kind of give me a feed on their thought process. So I kind of build on top of that. That's what I gauge towards before I jump into the technical things because technical things, I think, you know, you can make up. But if you don't have the right character and personality, it becomes extremely difficult to survive in a cross-functional kind of role.

Sheikh Shuvo:
Yeah. Awesome. That's all great advice there. Along those lines, say you were a university student today, just graduating with a CS master's and interested in exploring the world of AI, but often confused and confounded by the wider range of opportunities available everywhere. What type of direction would you give in helping someone pick that first job as they think about a longer career in AI?

Vinesh Sukumar:
Well, from my experience, I've learned that in your first job, try to develop as much breadth as possible. And don't try to be a specialist in your first job, because the moment you develop specialty, you end up being pigeonholed into doing one thing and one thing only. Right. So in the early phases of your career, try to explore. It's okay to fail. It's okay to fall. But at least you have a different perspective on different things, what not to do, or how to get things done becomes the most important thing.

So I would say, you know, try different sets of stuff. My expectation at that point in time is you've got, you know, less of any family burden at that point, or any kind of, you know, obligations. Usually, when fresh grads come out of college, like me, it's like a free bird, you know, getting that nice, empty canvas. What do I do kind of stuff.

So, you know, if I had an opportunity to go back and change myself as a younger person, I would have learned more languages, for example, try to get a little bit more understanding of statistics, try to understand biology a bit more because those things fundamentally help you in the future.

Right. And also have a passion for a different skill. The skill could be in music, sports, whatever it is, harness that. Because what I have seen in my personal life is you cannot be a bookworm and expect miracles. You've got to have your mind focused on something else. And only then, once you get refreshed, you can come back to your technical aspects. So, you know, I kind of have a passion towards that as well. And I'm pretty sure you will shine.

Sheikh Shuvo:
Awesome. Awesome. Yeah. Speaking of colleges, I saw in your background that you're affiliated with a couple of colleges as a member of the industry advisory board. Could you talk a bit more about what you do in that capacity?

Vinesh Sukumar:
Sure. I mean, the expectation at this point in time is to have education sessions with some of the students expecting to graduate from the school and give them guidance on what the industry is like, you know, because they're kind of entering the field for the very first time. How they could be successful in that space is one aspect.

The second aspect is more towards the university: how can we influence the academics, the physical collateral that goes into the books or the entire curriculum, so that it's much more meaningful for them when they come out. Right. It's not stale by any means. Yeah. So that way it's, I guess, bidirectional. I learn from the students what they are expecting, and we kind of give it back to the university and the students, what they should be doing so that they can be successful when they come out from universities.

Sheikh Shuvo:
Awesome. Cool. Well, and the very last question I have for you, Vinesh, is if one of our listeners is interested in getting in touch with you, what would be the best way to reach out?

Vinesh Sukumar:
They can absolutely shoot me an email. I think all my information is mostly public, I believe, on LinkedIn. So, I kind of maintain a pretty strong portfolio there. And they can absolutely reach out, contact me, and I'll be glad to help out.

Sheikh Shuvo:
Awesome, Vinesh. Well, this has been super helpful with lots of practical advice. Thanks again for being a guest on Humans of AI.

Vinesh Sukumar:
Thank you much, Sheikh. And definitely had a lovely time talking to you today.