top of page
  • Writer's pictureKen Jee

What You Can Learn From Teaching Data Science (Ajay Halthor CodeEmporium) - KNN Ep. 89

Updated: Mar 22, 2022

Today I had the pleasure of interviewing Ajay Halthor. Ajay is a Data Scientist with an interesting life story. Born in Los Angeles, he spent half his life in the States and the other half India. He dabbles in all that is data during the weekdays. And on weekends, you'll find him on a mountain. Today we’ll talk about his journey from a Computer Science major in India to Data Scientist and YouTube content creator & educator he is today.



[00:00:00] Ajay: Concepts like in machine learning and data science, let's be real. It's not super simple. There are very abstract concepts here that, you know, it's just really hard for people getting into the field to even understand and wrap their heads around. And that's totally fine, but it just makes it so, like it makes as a beginner. If I was watching all this, I would feel like, Oh, wow, I'm never going to really get there because I don't understand anything that's going on. And so I kind of wanted to target this in a way of like, How do we explain a concept without losing information.

[00:00:38] Ken: Today, I had the pleasure of interviewing Ajay Halthor. Ajay is a data scientist with an interesting life story. Born in Los Angeles, he spent half of his life in the States and the other half in India. He dabbles in all that his data during the weekdays and on weekends, you'll find him on a mountain. Today, we'll talk about his journey from a computer science major in India to a data scientist and YouTube content creator that he is today. I had an awesome time speaking with Ajay. I think you will really enjoy our episode. Ajay, thank you so much for coming on the Ken's Nearest Neighbors Podcast today. A huge fan. Yeah.

Excellent. I'm a huge fan of your YouTube channel. We've had some conversations offline and you're just a pleasure to speak to in general. So I'm happy that we could share our conversation, the deep, dark secrets that you, that you have. With all of the community here. And I think they'll learn quite a bit as well.

[00:01:29] Ajay: Yes, all of our secrets. It's going to be a wonderful time. I hope you all enjoy our journey for the next hour.

[00:01:36] Ken: Well, if you're watching, you can see Ajay's incredible shirt. I already mentioned that it was like, honestly, super sick. I want one, I guess it's not a shirt, it's more of a sweatshirt for those just listening, it has a bunch of trigonometry on it, which is pretty dope. And I've been doing a little bit of trigonometry myself recently. I've been using the manm package a little bit that three blue, one brown produced to create some visuals. Very difficult for me to use, cuz I suck at trigonometry, but it's a good reason for me to rehash and open up some textbooks.

So it's always fun. Yep, definitely. So I would imagine that your journey in data did not start with geometry or trig. Can you tell me a little bit how you first got interested in data? You know, was it just a pivotal thing that happened or was it sort of a slow progression?

[00:02:33] Ajay: That's a good question. So, okay. So let me actually like backtrack a little bit because you know I was when I came outta at this time, yes. Is when I came outta the womb, let's actually start there. I was born in Los Angeles. I always stayed here for about 11 years. And I moved to India, stayed there for another 11 years and now moved back to the States.

So let's see. So from a career standpoint, my first exposure to data would be somewhere around the third year of my undergraduate degree. Now until then, you know, I majored in computer science. And so during that time...

[00:03:08] Ken: So did you go to school in the US or India.

[00:03:11] Ajay: In India. Okay. In India. I did my undergraduate degree there in computer science. And it was there that, you know, as soon as I getting into like computer scientists in introduced to the field, I was just so interested in like back-end development, front-end development, all this web stuff. And then after that, it was, you know, kind of doing things around networking.

As soon as you like learn it in school, you kind of wanna implement it in some way and practice too. And I'd be like super interested in those like hardcore computer sciencey technical concepts at the time. But then I think it was in my, yeah, my third year of undergrad where I was doing some research under a professor.

And this research actually was like, one of my first projects was a computer vision based project where you would try to look at brain images and predict what a person is looking at at that time where it's essentially going to be modeled as like a classification problem where, Oh, if you're looking at a pencil, activation patterns in your brain are in a certain way.

There are certain parts of your brain that are activated when you're looking at certain objects. And it was really cool to see how, you know, building this nice little fun model out of the box. I didn't know anything at the time. I was just like, you know, plug in place. I could learn as we, as we start.

Right. But it was so cool just to see that, wow. I actually have the power to read minds and literally that's how, that's how it was like my introduction to it to start. And you kind of need that introduction, right? It's gotta be like super fun. Flashy. Interesting. Even though it's not all, you know, not all dandelions and roses, but it was a really good had thrust into, it was a big eye opener for me.

It was like, Oh, this field conduced things. And you know, from there. I just started to do a little more predictive modeling machine learning projects. And I think I, one of my one of like probably a pivotal project for me that got me into like how expansive this field can really be, and not just, it's not just about modeling, but it's about, there's so many other aspects of solving a problem end to end was in my final year of undergrad where I actually built this system, which was like a speech to tech system for a language called Kannada. This language is like spoken by people in a specific state called Karnataka in India and not anywhere else. Well, unless you're from that state and you moved out but other than that, it's spoken by like a handful of people there. And I just wanted to try it out because I'm not very good at the language I can read. I can write, I could barely speak here and there, but not that great. So it was an opportunity to meet, learn about the language. I also needed to learn about like syntactical and phonetic patterns in that language in order to create this model.

And because I kind of use like hidden Markoff models to model all of these things. And it was really fun because I remember picking up a book about like just the literature of language itself, like how Kannada and all of these other languages, you know, there are like other Indian languages that are related to each other and how their sounds are produced, where their sounds come from from different parts of your mouth.

And it was just like a very fun, educational, immersive experience. Just learning about that, which I didn't even think I would learn going into the project initially. Right. And then it was like me trying to record my own data too, because you know, data scarcity, there there's hardly any data of that, especially this is like six years ago.

Maybe at this point there was no data online that I could find this from. So I had to literally, like I was in my bedroom. And I was like recording my microphone and on, you know, everybody at home, you know, if you have like a MacBook, you kind of, I didn't have any recording set. So you need to kind of put your mouth really close to the site of your MacBook in a very weird and awkward position that it's like, hello, you know, you're just like, you have to record like that.

I had to record like hundreds of sentences speaking like that. It was a challenge, but it was a super fun challenge and it turned out pretty good at the end. So I was really proud of that. And I think it was from there that it kind of pushed me into, to pursuing like data science as I did my master's degree, moving to the States.

And then here I am as a full-fledged data scientist. So I love it. Fun ride .

[00:07:31] Ken: Yeah. Well, so something I think is incredible about that story with you learning the language, is that sort of the data and the domain they converged. So you were able to learn something more about like the language itself. I mean, for me, I work mostly in sports and I get these observations that I see from the data where it just makes the sport. Makes so much more sense to me, even though I've played it for years, even though I've been in the domain. And I think that that's something really powerful that a lot of people don't realize is that understanding of the domain as well as the data is complimentary in really unique ways. And they lead to like these aha moments that you just didn't have access to before.

Or, and I find that happening to me all the time and I'm like, Oh my goodness, this is like so obvious. But I wouldn't have noticed that unless I was looking at the data or you see something in the real world, and then you look at the data and you're like, Oh, that isn't, that doesn't match up. That's crazy.

So, you know, that's just such this unique expansion that we can get from this domain of work and it's perfectly exemplified in your story.

[00:08:42] Ajay: Yeah. That is so true. And it, it just like also shows where the field of data science, you know, when I started it out, I always thought it was about, you know, straight up computer science, just, you know, you plug and play models like, and then like solely, and, you know, you get into these models, you realize there's so much math behind it, so, okay.

It's all about mathematics. And then, you know, you work more and more on these projects, like you mentioned, and also like what I mentioned too, and it, you realize it's a lot about solving problems instead. And so, because of it, it's like ability to solve problems and we can use data science and machine learning to solve problems that we have an end to end holistic experience of, of just how everything just flushes in together.

And along the journey, you just learn so many things that you just didn't expect to, which is which I think is like fantastic. And there's always like room for or growth. Great. It's also a great field of interest because of that too.

[00:09:36] Ken: That's awesome. And so, you know, I didn't realize this about your story.

So you started in the U.S. You moved to India and you came back. What were the, what were some of the experiences like, how were they different? What was, what was that journey like? I don't think I've ever had someone on the podcast actually known anyone that's sort of gone like that direction. It's usually India to the U.S.

I think Chanin the Data Professor went to school in the U.S. and then moved to Thailand. But it's, you know, I'm very interested in like culturally and education. What were the differences and what was that experience like?

[00:10:10] Ajay: Well, yeah, that's actually, that's a great question that I have been more recently trying to, you know, disentangle that in my head too, of like, you know, what has the shift to India and back you know, done to me when I look at other people around too, on the plus side, I do see that, you know, I have gained a lot of perspective of just how people live their lives, because I've met very different kinds of people, just, you know, living life, literally, you know, if you were like an average person, let's just say they were, they're born, raised in a town you would, you're more likely going to be sticking with the, those group of people.

You'll have your little clique, you know, maybe in high school, middle school, high school that you kind of still are friends with, even till this day. Whereas like for me in that case, well, I have made friends, but then, you know, every single year, even when I was in India, I'd had to move a lot. So it's like, I remember fifth grade, I was in America.

I was in one school, sixth grade. I moved to another school in India, seventh grade, another school in India, eighth grade, another school in India. So it's like four years in a row, four schools like that. Right. And because of that, you get to meet a lot of people, you have experiences and you understand like all kinds of exist at a very young age, which is a great positive.

But of course, you know, that comes with a negative aspect of like, you know, there is no sense of stability. There's like life just keeps changing for you. Again, positive you adapt. Well, negative is like, when is this? Do you see an end in sight to it? You know, is it ever going to be under your control?

And especially, you know, this I'm talking about like grade school, right? So this was, I was a teenager at the time. Where, you know, it gets pretty rough as a teenager in general, moving around so much. But you know, as in, when I grew to become an adult, I could make more decisions on my own. Also like financially too.

I am, I can support myself as you know, I grew older. And so, because I could make more decisions on my own, I could also choose when and when not to move certain locations and also associate myself with the people who I choose and choose not to. And over time, I think like as an adult, I can look back and say, Wow. Okay. So these experience actually did help me gain some perspective. It did help me understand individuals at a very early age, which I don't know if I could have done it if I were in a single city for my entire life. You know? So yeah, overall that's, that's kind of my, my, my. Overall take that. I am still trying to disentangle till this day.

You know, I think about it every now and then just like, Hmm, what ifs? What has it done for me? What hasn't it done for me? What am I missing? You know, all those like self therapy questions that you kind of, you kind of go through in your head when you're just like sitting alone in your couch and just like looking outside. But it's definitely fun to think. Think about. Yeah.

[00:13:04] Ken: Well, it's funny in my last podcast, I talked with Demetrios who is start of the ML Ops community. And we, we had this incredible discussion about getting out of your own comfort zone, right. And something that not quite completely similar, but somewhat similar to yourself is that after like for college, I went away from my hometown after college.

I lived a little bit around where I grew up. And then I moved to a completely different city. And those were some of the most valuable experiences that I had. Like first I was able to see a different culture. I was able to like meet new friends. I realized that the world was bigger than just like my high school friends and my college friends, right.

To me. Like having these different chapters helped me develop as a person and really understand what I liked and what I didn't like. Right. If you're stuck in the same hometown, you go to school nearby, you do all these things. I mean, there's nothing wrong with that. You might enjoy that. Exactly. The world is small, right?

Like what I found from traveling, from doing all these things, the world's a huge place. Like experiment try you learn that. Like there's such cool stuff out there. There are things you might like that you didn't like. You know, the other thing is that. I guess with these experiences. You develop this very different sense of self because everywhere you go, you get to choose who you are in that place.

Right. So I moved from like Virginia to Chicago. I got to leave all the things I didn't like about myself in Virginia. And I got to bring and be the person who I wanted to be as I changed and moved. And I think that that's a very powerful thing. Like there's nothing that can help you make a personal change.

Like. Reorganization of your space or moving to a different space or, or like going somewhere where you have a clean slate. I mean, I did that in college. I transferred schools. I had like a two, five GPA I transferred and I became legitimately a straight a student. Right. I was able to reinvent myself because no one knew better.

Right. Yeah. The school didn't know better. And then when I applied for jobs, they just saw the. The basically 4.0 GPA and they didn't care about the previous one. So I think that there's some very powerful stuff like psychologically, and, you know, obviously there are the bad parts. Like I don't feel like I have a ton of friends that I'm like, wow, I'm like really close to this person.

This is my lifelong friend. We think about things the same way. But I don't think I would trade the experiences that I've had for something like that, because I could still make

[00:15:35] Ajay: those right. Definitely a hundred percent agree. And also like something I've realized, you know, it's just, if you were to look at people of a similar age group, it's always, it kind of like, seems like you would want to compare yourself to people around you, but I feel like as you get older and older you realize like how more different you are from just even people of your age group who are, who may even be in your same town. Like, like you and me too. Like, we've lived completely different lives as well, but here we are today and it's just so hard. You can't draw comparison so easily, even though we're both data scientists, it's like our paths and di like the that's probably like the only thing that I can think about that's like common between us.

But other than that, it's like, We're we've lived such different lives, but we've lived and learned, had lessons along the way. And now here we are like, hopefully like trying to make, do with like what we have using our past experiences, good or bad as lessons and just, you know, forging paths forward both professionally and personally in our, in our lives.

[00:16:35] Ken: So, well, you know, it's interesting. I'm 32. Now my birthday was a couple weeks ago. Happy birthday. Thank you. And a bunch of my, you know, a bunch of my friends are, are married. They're having kids they're doing this and that. And like, I don't feel like I'm ready for that. I feel like I have so much more to do before that next step, but I, you know, also like you look at me, I like think I like feel, and I look physically younger...

But, you know, it's like one of these things where like you know, like their journeys are not mine and like, yes, like, you know, it's nice to find someone it's nice to do those types of things, but like my path, that's not something that I'm ready for yet.

Do I want to have those things one day? Absolutely. But I wanna set the stage so that I can have a family and provide for them and do whatever it is in the best possible way after I've gotten all of my, my stupid entrepreneurial things. Done with, but you know, my body seems to be in line with that soon to be cooperating with me.

And like, my lifestyle seems to be in line with me doing that at a later date. So. Why not live that journey and maximize on, on who I am. Right.

[00:17:51] Ajay: That is so true. That's so good. Because like, yeah, it it's really easy to just jump into this bandwagon of like, Oh, everybody around me is getting married, so I should be getting married too.

Like, I also kind of feel that as well, like I'm 26 right now. And at this time there's like a lot of my peers are probably in the first wave. They're yeah, yeah. Again, in the first wave of like everybody around me is getting married and then, you know, this is like when they start prodding about, so what about your relationship status?

Mm. Are you seeing anybody? Hmm, what's going on there? You know, you get all of that, like from everybody around you and you're like, it's like, you wanna record something and just, just like putting your phone and be like, no, no, not yet. I wil let you know.

[00:18:34] Ken: I would love to collaborate on a skit with you of like you're on a date with a data scientist.

I've seen those, like you're on a date with an investment banker and they're just like, Oh, you work in marketing. Let me tell you about fun. Like, I dunno. I think that that would be such a fun. Anyone that's rich but I digress. So. You know, I'm also interested in the education aspect. So you went to college in India.

What was, were there differences in like university there versus your grad degree here? I think that that's something that I dunno, I'm fairly interested in. I've never gone to school outside of the U.S.

[00:19:12] Ajay: Yeah. That's a, that's also a great question. So I well, I don't really have a, I only did like undergraduate there and then grad school here, so I can kind of speak to that.

I do know that like in India, when you do have like the education system, there seems to be more, more theoretical in nature. Like you do, you do have a lot of, you know, found fundamental, foundational knowledge that you do learn from textbooks. That's really good. I think, you know, obviously like there is this aspect of like how practical is it and also how are you encouraged to explore these new avenues? Like sure. Like, let's say that you are specifically interested in like mathematics or computer science, but also how are, are you empowered to actually do the projects that you want to do that are practical and that may not be like, so, you know, so straightforward to your degree or contributing to your degree, right?

Because there might be like in India, I know, like I had to go out of my way to kind of look for like, do research under certain professors that I wanted to, I had to participate. I also did like. I participate in certain conferences. And I tried to do some of my own research, present them as well, but these were not really affiliated with my own university.

And it's not necessarily things that a lot of people around me did. So I was kind of like the lone Wolf doing my own thing. Nothing that there's obviously nothing wrong with with like just doing the straightforward path. In fact, there are so many very successful, very intelligent people. My peers, who, who graduat with me in undergrad as well, and they're doing fantastic right now.

And I, you know, I still keep in contact with them, but, you know, I do see as a and like right here when I first came here, well, Returned back. I guess I did notice that there were so many people who I met, who were not from India, who like they had like so many other things going on. Like when I was a grad student here, I went to the university of Southern California, USC.

It was there that I made some friends in undergrad and I saw, wow, they're doing so many like extra, I guess, like somewhat extracurricular activities, but also kind of related to their field, kind of not. But they're doing like so much in terms of like diversity from a diversity standpoint, which is that's, that's really, that was really interesting to see because you know, in India it's like you do a lot of things for your degree.

You do things to get your degree. Anything that is tantamount to that degree, it is always like looked up upon no matter what, but anything that's on the aside, it's like do you have to do it? I don't know, but there is still like a lot of projects like that, you know, instead of doing in lieu of those like diverse projects, they try to go more deep into that one thing.

So it's kind of like a more depth approach in India. Whereas here it's like, you have a lot more expansive options too. Maybe a lot of this is also cultural because, you know, If you look at people from India, they tend to be like, a lot of them are in the engineering or any of these STEM fields.

Right. Whereas here it's like you're out of college, straight outta college. You have a lot of people almost across the board. Right. So I think that's also reflected in society as well.

[00:22:37] Ken: Interesting. So like like practitioner versus more of the liberal arts philosophy? Yeah. You know, I think over time I've become less enthralled with the liberal arts methodology to be perfectly honest.

I mean, I went to multiple, at least multiple liberal arts schools, and I found that I took a lot of classes that I just wasn't interested in. I might have been, well, I actually, wasn't more well rounded. I like hated history. I hated a lot of the writing courses I took and it wasn't until after, when I wasn't required to take 'em that I gained appreciation for those types of things.

Right. Exactly. I digress. I think it's interesting, you know, culturally that you had those experiences and I'm wondering, did that shape how you viewed education in general? I mean, you know, you were you'd to me offline that if you weren't a data scientist, you'd love to be an educator in some way. Do you think that those experiences that you had across different ways of learning might have contributed to that?

[00:23:40] Ajay: Actually, that is also a great point. I know I've been saying great questions, great points. But yeah, you do bring up a great point because like, these are things I have been thinking about too. Again, more recently though, like I'm looking at my life. So like, in retrospect of like so many different types of experiences and how I am now and how, you know, those different experiences have, have changed or culminated into to how I teach. There is like that the, I didn't actually, first of all, become an educator on YouTube until I came to the States. It was only during my, my master's program here that I actually started my YouTube channel. But I guess the reason for that is less of cultural differences like between like how my past was and how it was at that time.

And it was more of like, I don't, well, first of all, I don't see great resources that really go into depth into like really deep technical knowledge to an extent that I am satisfied that is just freely available online. And I thought, well, we can change that. You know, I just wanted to like, and in the very beginning I just wanted to learn, right.

Learning was my main objective. And the more you teach someone, the more you can actually learn yourself. Because you realize like, Oh, you think, you know a concept it's like, Oh yeah, support effect. Oh yeah, Support Vector Machines. I totally know that. And then when you try to actually write down things about like, what, you know, you realize like the holes in your knowledge and this kind of ties into at the time I'd kind of like I was really into to Andrew GoGy, which is like adult learning. And I wanted to know what was the best way to teach adults and teach oneself in move that, because I was, I became an adult too. Right. So I did, I totally became an adult. I swear I'm a above 21. I have an ID. Okay. So anyways, What I did for, you know, to cater this was like, first, let me, let me, let me talk about like this technique called the Fineman technique. It's basically a methodology that you can use to learn anything. So essentially you have a very complicated concept, right? Or any concept at all. If you wanna learn that concept, the best way is to take a sheet of paper and basically explain that concept in the simplest of words that you possibly can without using any, any kind of jargon.

And once you write, once you write a draft of your explanation, You go back to it and just make sure that you're really not using any jargon that could throw people off. Right. If you're able to break down the entire explanation without jargon then that's great. You've kind of understood this topic, which is fantastic.

But if you haven't, well, you realize there are so many holes. And when I did this for a lot of these machine learning concepts, I could, I saw, okay, there are things I don't know. And there are things I need to look up. There are things I, there like a lot of wood to be chopped. There is work to be done. And so I kind of I was, and I was also studying for these like four tests. Like at the time I was still in grad school. So while I was like making, I wanted to make these videos more as like a study guide for myself too. And so I would kind of like start writing out mathematical derivations and explanations and how I knew.

And when I started uploading these again, it was mostly to teach mostly to teach myself these are like older videos, like back in those days, like four or five years old. And they became popular, but a lot of it was good traction, but a lot of it was also like, I can't understand anything you're saying was like my, my comments and I was like, you can't understand what I'm saying?

And I looked at my videos again, like a year later. And then I realized I'm watching my video and I'm like, I know what I'm talking about, but even I can't actually follow completely what I'm saying. And that's when it started clicking. I was like, I'm doing it all wrong. I'm not, I'm just like talking.

I'm just like, kind of like talking jargon on top of jargon without completely knowing what I'm saying. And so let me actually take a step. And try to write this, these scripts in a way that more people can understand too, because it's not just for people. It's also for myself. If you can simply, you know, just break these down.

And so over time, if you kind of like, look at my channel videos, I feel like I'm, I was, it's a slow progress, but I've been slowly trying to get better and better at just not throwing terms on the screen, throwing jargon on the screen and explaining jargon with jargon. Yeah, that was, I think like of very big, the, like, it wasn't a pivotal moment, but it's like a gradual pivotal you know, Change eventually over time that I could see a difference in how I taught, you know, three years ago to how I taught teach now.

Yeah, another aspect to that is also like something I learned along the way, which is like that Fineman technique for learning is fine. But for teaching, I don't feel like it was like complete because it's easy to teach a concept to like, I don't know, let's say a five, let's say you're explaining a concept to a fifth grader versus like an 11th grader.

Right. So you know, have you, I'm not sure if you've all seen these videos of like explaining concepts that five levels like explaining quantum computing at five levels. Right.

[00:28:56] Ken: I want to do that for machine learning, but someone beat me to it.

[00:29:01] Ajay: You know, you can always do it, man. This it's never too late. But when I watch these videos, I feel like they're really cool, but I feel like they're the point though, when you're kind of trying to convey information, is that it would be even better if you were not losing information when you're conveying it to somebody who is younger. Clearly, it's very clear in those videos when they're explaining like the blockchain, they'll only explain like basically in two sentences and they lose a lot of information when they're explaining the blockchain a little too much information in my opinion too.

But I wanted to explain like, concepts like in machine learning and data science, let's be real. It's not super simple. There are very abstract concepts here that, you know, it's just really hard for people getting into the field to even understand and wrap their heads around. And that's totally fine, but it just makes it so, like it makes as a beginner. If I was watching all this, I would feel like, Oh, wow, I'm never going to really get there because I don't understand anything that's going on. And so I kind of wanted to target this in a way of like, how do we explain a concept without losing information? Right. So in looking at it in this lens, you can see how explaining calculus or like the blockchain to a fifth grade to a five year old or a fifth grader or yeah, a fifth grader is so much more difficult now than explaining the blockchain to an 11th grader.

Because they have like fifth graders and 11th graders just have like foundational knowledge that, you know, you can kind of build on top of. Right. And then it, it kind of like gets back into like this entire, how is education done in school? Right. Let's start with this. I know I'm probably jumping topics here a little bit, but it'll tie in.

I swear. Cool. So, you know, in like traditional education, right? So let's say you're like a fifth grade student. You're a student in the fifth grade. Your student, the students themselves, they will be, they'll have the level of education of a fourth grader, but the level of the concepts that you are teaching in the fifth grade are fifth grade level of concepts.

So, let's see. What is the difference between the level of the audience and the level of your teaching or the concept? That's like five minus four, which is like one, it's a difference of one. And let's say, okay, next year you move on to the fifth grade. And you have a, you have like, sorry, you move on to the sixth grade right now.

And as a sixth grader, you have the level and, you know, the ideas or the knowledge of a fifth grader. But the concepts now that you're learning is the sixth grade. And so there's still that difference of one in that con in, you know, what you're learning and like what the concept is versus what your background knowledge is.

It's easier to pick up. And this is how the traditional education system kind of. Kind of works like one year by year you're you're kind of like stepping up concepts little by little over time until you hit like 12th grade. Right. But now with online learning though, that becomes so much more different because the level of your audience is it could be anybody.

Literally. You're not just teaching a spec, not everybody in your audience is going to be, be the same level, especially with adults. Right. This also ties back into the fact that adults are so different. Like, they all have different backgrounds. They all have different experiences. They've lived different lives and they've culminated all that knowledge to become the people they are today.

So the level of your audience is variable and the level of your concept, which is like, let's say if I was making a video on transformer neural networks, the level of your concept is like fixed technical way. The concept is fixed, but the level of your audience is variable. But now bridging that gap between the level of your concept and the love of your audience.

It's much tougher because, well, first of all, if the difference between the concept level and your audience level is a lot, it's gonna be way harder to, you need to bridge that. That's really hard for you to explain. And the fact that there is a variability in it also makes it even more difficult to explain, because there's gonna be a lot of people who are gonna be like, this was a great video and there's gonna be a lot of other people who are like, I didn't understand anything.

Right. Because of that. And this kind of led me to think of like, okay, let's try to spice up my online content even more. And I came up with well I came up with as if I'm like this professor and yeah, I'm not a, no, I'm just like a random dude in a house recording YouTube videos. But I thought of trying to, I wanted to try a strategy of like a multiple past explanation of like, where the first I'm gonna explain, let's say transformer neural networks at a very high level. Just start to finish though. Very high level, just like run through like the AR the foundation of the architecture, treating it as a black box and just say what it does, for example. And then on the next pass, I wanted to go through actual details of like, okay, how are let's talk about the math here of like, okay, a vector is going into these, these, this little machine learning model.

And it's, you know, going to be encoded in some way. And that's how I kind of explain it. And then on the final pass, I'll try to go even more details into just. Trying to patch things up there in the explanation and doing this approach. It's like, well, a person now who is, you know, who who's really new to the field, but interested in transforming neural networks, which is a hype topic, they'll feel satisfied out for the first few minutes.

And they feel like they've learned something because they learned it end to end within the first pass of explanation. And then people who were really into the field, like much deeper into the field, they can just watch the entire video because eventually, you know, they kind of just like start understanding more and more details.

Yeah, this is, I feel like this is kind of like the strategy I've been trying to use for a while in delivering content. And it is, it is exceptionally harder to do this though. If you were writing a book because like, if you're writing a book well, yeah, because like books by nature, they're chapter by chapter, right?

So if you were to all, all you computer, computer people out there reading a book is like, kind of like a depth first search kind of thing, where every chapter, it talks, it starts with a topic and goes really deep into it. And then you go into the next chapter. Starts with the topic goes really deep into it, but clearly the way that I'm trying to formulate my videos is more of like a breath search or a breath traversal, where you're trying to do the, like a specific concept at like high level.

And then, then you go deeper all the way through. Try to explain it then, and then go another level deeper, try to explain it all the way through. And I felt like that this methodology, at least for adults, I believe that it works a little better. But again, I'm still trying to figure this out myself.

So it's always a fun game. The cool thing is that I'm always learning about this, and I've been thinking about this for years as my content for technical education has been evolving over time. And the cool thing is there's always much to learn, but the uncool thing is that there's always too much to learn but it it's fine.

[00:35:56] Ken: Well, I think that there's a huge problem with traditional education in that when you're in a classroom, Everyone is forced to learn at the same speed and at the same level. Yep. Right. And in an optimal scenario, people would be adequately challenged, but not like overly challenged or, or not challenged enough when they're learning something and content now, YouTube, any of these places, they allow you to match content to the difficulty that people are interested in.

Something that I really like. And I do a lot of is like there's underlying concepts. And I can link a video that I've made on those underlying concepts that explain them better. If someone knows it, they can keep going. If someone wants to stop and watch that other video, they can, right. Josh starer does this really, really well on his stat quest channel, right?

There's like, well, you know, there's this other video on gradient descent, which is really important here, but if you're familiar with gradient dissent, just skip it it, and you can keep on with what we're doing here and the just modular nature of content content, and this choose your own adventure style.

I think is something that is unbelievably valuable as we continue to think about how people are gonna learn in the future. I just, you know, I did a lot of formal education and I saw what was wrong with a lot of formal education, you know, early in my career, it didn't work well for me because I was behind the curve.

I was really slow. And then later in my career, you know, in latent college and in grad school, I was learning it about the right level. I wasn't way ahead. I wasn't way behind. I was like a. An average learner, but because this, the rate at which I was learning was average, I excelled better than people who learned it quickly because they were just bored in class.

I was super engaged. And so like, it's kind of backwards that I, like, I like, you know, backhanded my way into becoming a really good student, but I was one of the rare people that I think education systems and formal education worked well for, because I was like specifically at the learning rates. That that was optimal for the equation at hand, we could go down with learning rates and get real nerdy, but I'm gonna...

but so I'm interested a bit more in just like your progression of iterating on content in the channel. Right. You've talked about how your you've You know, you started and your, your longer term goal is to make these videos and teach things in a way that they're digestible and they're matching the specific audience.

But how do you implement that? How, you know, you you've told me that you've gone back and made remade a couple videos. and what, what would you like to be able to tell AJ who was just started making videos about, about your content and about how things would progress?

[00:38:50] Ajay: Do you understand everything you're saying in your explanations?

Well, to be honest, I actually thought I did at the time, like Yeah, I think that when I a good, a good example, I think for answering that question would be a foundational concept of logistical logistic regression, where I think I've made multiple videos on this. If I look at my like very first videos, it kind of looked like a slide deck where it had the right information in there, but it was just very boring to watch, you know, and very boring to digest where.

Sure. Maybe it would be, it would pass as like, if I wanted to give like a, some, some lecture, but like not everybody is super enthusiastic about it. Right. And honestly, like, even if you are interested in it, you know, sometimes you'd be a little like, okay, is this, is this how I think it's working? Because there's like so many, I use a lot of words and text on.

I slide like presentation at the time. And then when I tried to redo that, I actually used manm. Fun, fun, fun tool that you mentioned before by grant Anderson from three, three blue one brown great tool. And I tried to visualize like, exactly how, like the sigmoid would look for like two dimension functions and three dimensional laws as well.

And that, like I learned so much just by. You know, creating that myself. And I think it was also a very good visual aid and learning tool for other people who are watching it. And regardless of their level of knowledge of the field, like, Hey, this is what it looks like. This is actually what's happening.

Even if you are like advanced in the field, you'll be like, Oh, that's interesting. And if you were getting into the old, you'll still be like, Oh, that's interesting. And I think like at least starting to playground with like those foundational concepts into how I deliver content is, is how I. Kind of have tried to evolve in the past, like I've done the same, like, and to that, and like I've also done the same for like very, very seemingly simpler concepts.

Like what is a loss function? What are optimizers? What are what is batch normalization? You know, these, these small pockets of like foundational knowledge and information where, you know, it, it's better explained with jokes sometimes, and also with actual serious content too, because you need to kind of add some element of fun to it, you know?

Like I try a Anthropomorphize the concept of optimizers where I think I've created like one of the comedy shorts on this too, where I treat optimizers like hikers. So if you treat like every optimization function as a hiker, and then the loss function that you're traversing as like a canyon that's to be hiked like, Oh, our goal is now to hike to the lowest point of that canyon and who can do it better, who can do it faster.

And then I have like, you have conversations between like SoCast gradient descent versus like Adam and being like what, and trying to belitle the other. And it's just, you know, it just creates a fun dynamic, but. From a comedy perspective. That's great. But also from like even a mathematical perspective, you can see how one changes into another and how it can become, you know, like small mathematical modifications of like astic, gradient descent can lead to like a mini batch gradient dissent, which can lead to, you know, add adult to Adam, Adam, and all these other.

Optimizers that we, that we kind of have that that gets a little technical, but the core idea is that, you know, just making some of these concepts more relatable when you're explaining it really goes a long way and not just to people who are getting into the field, but even for people who, who might know the fields.

Pretty well, and it's just like another perspective. Again, it all comes back to perspective too. You can look at the same thing in so many different ways. And I think your knowledge will only expand by just introducing yourself with different, different ways of looking at the same thing.

[00:42:50] Ken: I really like that. I think something that I learned early on, and this is what I recommend to people. It might not work for everyone, but I've loved learning programming first and being able to apply the math using programming, right. That made it very visible to me. It made it tangible. It made, made it not theoretical. And when I got into theoretical math, that was intimidating.

But as long as I knew I could code it up and see in tri practical examples. That's when it really made a revolutionary, like change in my understanding. And so I think what you're, you know what you're describing, you're getting the same thing when you visualize these things, you're like, wow, I understand this so much better.

Like, it just makes so much more sense when you implement it. And I think that's a really. Important thing, not just for teaching, but for learning is you get to experience with these different ways and these different styles and they can be UN unbelievably valuable in your own personal development. I am interested in how you evaluate your videos and your performance.

Is it just the clarity? What are the things that you like? Optimize on or that you want to improve upon as you go, like, what are the metrics that you care about with your videos and not necessarily like viewership or things like that, but like, yeah. What do you, when you look at a video, you're like, how, how could Improve this?

[00:44:08] Ajay: Yeah. So that is like an ongoing process of research for me, because typically what I do is when I record a video I try to premier my videos. So I'm able to watch my videos with my audience when they launch. And this kind of is like an accountability standpoint. Like I will watch my video. You at least once with you all while you're there too, to make sure that.

I just wanted to make sure, like in the event that I do say some things that require clarification, I can clarify it on the spot and more so even for myself where it's for a self-reflection of like, okay, I'm watching my content now. And I see that, you know, there's some points of explanation now that are not as clear to me.

Because, you know what, when you're creating these videos and probably you do the same thing, it's not like you, sometimes I just create the video and upload it where it's still fresh in my mind, even when I rewatch it. But there's also a lot of times where I have videos from the past. Like I recorded maybe a month ago, I did all the research.

I recorded a video and I watch it like in premier, like one month later and I'd be watching it now with kind of a not a complete slate of the entire picture in my head. And so that gives me a good chance to evaluate okay. How would a person just coming into this content view? My own content. Because I don't have anybody in like a team to just like peer review my content, just to make sure that everything kind of like makes sense in the entire, you know, it flows well.

But if I do watch like my video off for like a month or so, and I see, okay, I get what I'm saying, but not entirely. Then that's like a signal for me to be like, okay, this is like a point where I can potentially improve an explanation. What went wrong here? If it did. Typically, it's really hard to say there's, there's no like wrong or right.

It's very hard to binary categorize these things. Like it's not like it's wrong. It is correct. It's just like, it doesn't resonate very well or it doesn't click very well or it's just, you know, it, it was just presented in a slightly bland way where it could have been presented in a slightly different way.

Maybe if I, you use like bullet points, why did I use bullet points instead of like an actual graphic here? If it could have been made or done pretty easily because, you know, visuals do help you know, just evaluating myself kind of like that, and just gradually making tweaks over time in future explanations so that, you know, they just become better.

So it's never, again, this is nothing, you know, nothing that I've seen has ever become like, aha, this is the way to do it, but it's just like a gradual progression over time where. I'm able to kind of learn lessons from just small, small quirks that I kind of see in my videos that I wanna iron out and just over time, it just becomes more and more refined.

[00:46:53] Ken: Awesome. Well, let's change gears just a little bit. Something I'm interested in and you've expressed that you enjoy cool is adding the new comedy skits using shorts. I've watched a couple of yours. I quite enjoy those. What's been the inspiration or sort of the onus for those.

[00:47:10] Ajay: Well, typically I'm sure that you every, like when we're talking very professionally and when I'm also delivering a concept, You know, it's very easy to get into a monotone voice.

And also like for the viewer who doesn't see my face very often, I'm only presenting a concept like this and I'm talking about vectors and you know, like whatever the embedding space and how things work in very technical concepts, very fun stuff. But, you know, it's really hard to kind of like inject my own personality and flare into those videos, unless I'm actually doing something a little funky in those videos, which cannot be for an extended period period of time, because after all it is supposed to be an educational, technical video.

So you can't like you can't spend like of a 10 minute video, you can't spend like five, just, you know, Hey, let's just talk about hikers and optimizers and you know, how they talk to each other like this? No, you can't do that. Right. It's you gotta be, you gotta be a little more like direct about it so that the information is very clear, but you can try to spice things up so that they're not boring, but even then it's like my personality.

It's like, I pop a lot. I was like, we, you know, I, yeah, I pop basically. So I wanna kind of. Interject that somehow in my content and shorts was actually a really good way to do this because, you know, it's really hard to convey huge technical pieces of information within like a minute video, but it is easier to convey like just, just visuals of just like, you know, you acting funny in a way that kind of, kind of relates to to a technical concept where, you know, data science, like anybody who's in the field will just look at it and be like, ha ha.

You know, and then they just like it, you know? And it's like, Oh, okay. That's some, that's some exposure. And I also kind of express my own my own comedic ins here, which I've always wanted to sew for so long. Just get out. But yes. Yeah. I did not turn into yeah, no, no SNE on here. No, no, I'm just, yeah, that's me.

So, yeah, it's just like another perspective of, of just my own, the facet of my personality. Although in the end I do see YouTube right now. It's like, Being a reflection of my brand. So I do like to keep it professional and that's how I like to do it moving forward too. But I think like in isolation, some of these skits are just, they're fun to do, and they're not really harming anyone or anything.

And it's just like a fun outlet, a creative outlet for me because I get to sit down just like create these little skits and then, you know, act them. Do some little direction, micro direction, whatever that requires like, Oh no. Put the camera here. Put the camera there in my little, my little phone over here.

It's like a nice little Android phone. That's like three years old, fun, fun cameras there. You know, cameras do wild things when they're aged, but yeah.

[00:49:53] Ken: We can find your most creative content on only fans. So.

[00:49:57] Ajay: Of course. Yeah. I hope you're watching everyone.

[00:50:03] Ken: Incredible. So the last thing I want to touch on, which I think is really cool is, and you've described hiking and those types of things when talking about optimizers, but.

What is, what is hiking and getting outdoors done for you in your life? You've talked about how meaningful that is and how much you enjoy it. I'd love to hear not just how it relates to your content, but how it relates to your enjoyment and and your leisure.

[00:50:30] Ajay: That's great. So hiking is my primary form of exercise.

I absolutely love hiking. I hike every weekend. Whenever, whenever I can, I would like to say rain or shine, but if it's raining, I don't wanna hike, but if it's shining, I'm definitely out there. I'm on the top of a mountain somewhere. You will call me in the middle of the day and I'll be up there. these days, I think like, because of the pandemic, I've also gone into solo hiking a lot, because you know, at a time like this, maybe like a few years ago when I used to hike, which I still love to do, so I'd always need someone to be with me because like I felt like, Oh, it's kind of awkward to do these hikes alone.

but then over time, like I think in 2021, I started hiking alone and well, in the beginning I would be like, okay, I don't know what to do here. I put headphones on. I try to listen to music, but you know, it, it just kind of got boring doing that for some point. So I stopped using headphones and I just like, you know, just, I think one of the, one of the days I was high hiking one of my favorite peaks here, the peak is called Los.

Peak and Cal in lake Los Angeles. I'm not sure if any of you LA viewers have seen it, but it's a wonderful peak anyways, that peak there, there there's like an easy trail, a hard trail to get through a mountain range. I always love to go through the hard trail. You have to go like down this gutter and then up a mountain range.

And then up and down, it's like a hike along the ocean. It looks beautiful. And when I do that, there's like nobody there. And I could just hear the birds. I could hear the mountain. I can hear the sounds of nature. And I don't, I didn't, that was like, I think one of the first hikes where I just like, did not need headphones.

I didn't need to listen to anything. It was just a few hours of me alone. Just walking up, just alone with my own thoughts. Right. And all I could think about then was like, I respect the mountain, just respect the mountain and just. Thinking about how it kind of relates to, you know, other physical activities too.

Right. I don't look at the mountain as like a destination of like, okay, I gotta count my steps here and gotta make sure I, you know, do my steps for the end of the day, but more it's like, it's the journey that really, that really gives you know, that power. And it's like, that journey is very fulfilling.

It's like the view on the top. It's great. But some days it's cloudy and it, you know, you won't get a view, but you shouldn't be disheartened by it because. The journey to get, there was a really fulfilling and just, it was just a fun experience, you know, and that kind of makes my, my Saturdays or Sundays worthwhile too.

So yeah, you can also apply that to data science, I guess, in some ways, or even life in general, it's the journey that matters. You shouldn't really be too focused on there's no real like destination men after which everything is Dan lines and roses, and you shouldn't just kill yourself when trying to get to your goal, but you should also like, you know, live your life while you're doing so. Yeah.

[00:53:14] Ken: Well, I love the idea of respecting the mountain and respecting the journey. And if you do that and you're mindful through it, that's where the most enjoyment really comes. Right. You get this relief when you get to the top, but then you have to turn around and hike back down, or you have to hike up to the next peak.

Yeah. And there's always gonna be other peaks with views. Right. You can stop and take it in and enjoy it, but like, you're not gonna stay at the peak forever. You'll like die up there, right? Like no one climbs Everest and it's like, ah, moving in, right. Moving in. Right. So, I mean like data science, any, any journey it's like.

There isn't really a destination, right? It's always a continuation. And if you love the continuation, you love the ascent. You love the dissent, you love taking it in and you're not taking too many Instagram pictures at the top. I think that that's what makes it worthwhile. And if you have that approach, you're gonna stop looking for distractions or like looking for motivation and these types of things as you try and go, right.

You're motivated by the journey itself, which I mean, even for me, like, It's difficult sometimes, like you're hiking and you're like, man, it's pretty hot. You know, it's hot. Yeah. I'm halfway there, man. I'm tired. My feet are sore. That happens to everyone. But if you can appreciate that and you enjoy it as you go up, it makes the whole thing a lot, you know, it, you enjoy the experience, not just the highlights, which is that's so true, which is harder and harder these days.

So AJE super enjoyed this. How can people get a hold of you? How, how can people learn more? What is the best way to connect? What do you have coming up in the pipeline?

[00:54:49] Ajay: Wow. Ken, I thought you would never ask. Well, For all for all you lovely listeners out there. Thank you so much for listening in. And if you want to know more about technical contents in like machine learning, artificial intelligence, data science you can find me on code Emporium on YouTube.

That's probably the best way to see. To see my content. I do have a Discord server. It should be like in one of the links down in the, those videos, all stuff. Perfect. Thank you so much. I appreciate that, Ken. It will be there in this description too, if you wanna join in. But yeah, just, If you're interested in the field, please do follow along and I'm gonna do my best to keep improving in how I deliver content to you and everyone else.

And yeah. Things in the pipeline. Well, there are certain things that I'm definitely working on outside of YouTube that I will definitely announce once it's a little more refined too, but again, if you, if you do follow, follow along, I will keep you very closely posted with those updates and pretty soon too.

So lots of excitement here in 2022 cannot wait.

[00:55:57] Ken: Heck yes. Awesome. Well again, thank you so much for coming on. I really enjoyed the talk.

[00:56:02] Ajay: Perfect. Thank you, Ken. And thank you everyone for listening.

32 views0 comments


bottom of page