Fabriace_Mesidor_mixdown.mp3

Fabrice: [00:00:00] I find that coding and math. I mean, the entire machine learning models today are really complex. So pick something to have fun with. You don't have to be stuck with quickly coding enough to start working on something that you don't like. So you need to take a project that you will have fun. You just enjoy. And the second thing is, don't be scared of the challenge.

Harpreet: [00:00:36] Whatsup everybody, welcome to the artists Data Science Podcast, the only self-development podcast for Data scientists. You're going to learn from and be inspired by the people ideas and conversations that'll encourage creativity and innovation in yourself so that you can do the same for others. I also host open office hours you can register to attend by going to Bitly.com/adsoh forward slash a d s o h i. Look forward to seeing you all there. Let's ride this beat out into another awesome episode, and don't forget to subscribe to the show and leave a five star review.

Harpreet: [00:01:17] Our guest today is a data scientist and machine learning engineer who was born and raised in Port au Prince, Haiti. He's earned a bachelor's degree in quantitative economics and statistics and has over eight years of experience in the telecommunications industry. He's also studied law for four years. His hunger for knowledge and determination to turn data into insights has contributed to his most recent successes at the Digital PNG Limited, the biggest telecommunication company in Papua [00:02:00] New Guinea. He emigrated to New York in late 2019 and has since been focusing on his Data science career, where he's part of the business intelligence team at way better. A goal sharing network whose approach combines games, social reinforcement and financial incentives to motivate people to get better at everything. So please help me welcoming our guests today, Fabrice Mazdoor. Please, man, how's it going?

Fabrice: [00:02:28] Everything's good. Everything's good. Thanks again for inviting me. I'm so excited.

Harpreet: [00:02:33] I'm super pumped to have you on, man. I stumbled across your work where I was like, I peruse articles on Medium and things like that. You have this project that really caught my eye. That was like Data science and hip hop music. And I was like, Man, this is this is phenomenal. When to talk, to talk to this guy? He looks like he has some interesting ideas. But yeah, we're going to get into all of that and a little bit. But first, let's get to know you a little bit better. So talk to us about where you grew up and what it was like there.

Fabrice: [00:03:00] So I came from the Caribbean, so you just mentioned that I born and raised in Port au Prince, Haiti. So it's a really nice country with two weather. So now I'm in New York, but it's really different now. So basically, we have like beautiful beaches, so it's really warm country and going there. So it was always about education, so my parents always push us to go to school. It was like, So you have to go to school first. And probably the weekend you can meet some friends and watch some sports. And one of the things that really is the way that I am now. I think that is really the discipline that I get so many people from the Caribbean. They know that as parents, they are really strict. So you have to do this. So if you don't study so you don't get anywhere. So I think that that's really helped me in life now. And one of the challenges that we had in [00:04:00] Haiti was mainly because of the political turmoil that we used to have and there was a lot of natural disasters. So if you go to the Schumacher that we used to a lot of storms, so we have a big earthquake in 2019. So there was. So my salary in Haiti was like a mix of fun and I was where a lot of the political trouble or natural disaster, but it was really nice going there. So I really like the Caribbean. I've been to other countries in the Caribbean. I really love the weather and I mean, just the entire culture as well.

Harpreet: [00:04:31] Yeah, man, Caribbean sounds like so. For the record, right now, we're recording this in February. It's like the middle of winter. I'm pretty sure it's cold in New York. Like right now, right now here in Winnipeg, it is negative. Twenty six degrees Celsius with 10 mile per hour winds, so I wish I was in the Caribbean right now.

Fabrice: [00:04:51] Yeah, it's it's freezing. It's it's been snowing probably twice in the way. Yeah, it's really differently.

Harpreet: [00:04:58] So. So talk to us, I guess, like the walk us through how you you went from from Haiti to New York, right? So Haiti, you grew up so in high school when your high school age student, what did you think your future would look like? Do you think you'd end up in the middle of winter in York?

Fabrice: [00:05:16] Well, probably no. But I'm pretty sure that it will be related with Data. So let me just say when I started AIs my first year and I went to school because I wanted to an economic science, so that's something that I love and anything about the economic PDF. So that's why I wanted to do but the school that I went to. So we have statistics as well, and we have economics and planning. But at the end of my first day, I was like, I think that statistic is more for me. But at the same time is what I wanted to use economic policy to change the entire landscape in Haiti to to to be part of the [00:06:00] development. But one of the things that I realized it was like there was a lack of Data because if you want to make decisions so you need Data, you cannot just be as a as economists, you cannot just go say that, okay, that's what I want without any Data. So you need some Data. And there was a lot of data and all that. Probably this is where I really want to do my job. And I think that this. You know what I'm going to say? Forget about me, okay? I want to be. And I really enjoy doing a survey. And when I start doing econometric model, it was like, Oh wow, that's so much that I can do. And I find that when I was high school, I was like, OK, I think it is related to Data. I want to be part of it. And we didn't know if I will be in New York. But I felt that I already have a clear idea that I wanted to do a statistic and managing Data, playing with numbers in something that I really like to say make the numbers tough.

Harpreet: [00:07:03] Yeah, we got a very similar background in terms of the education because I studied economics and like statistics during undergrad years as well. So what do you think econometrics was like my favorite part of economics? Like, I just absolutely love that stuff. So what do you think? Like, do you like microeconomics or macroeconomics better? Which one do you do you prefer?

Fabrice: [00:07:24] I prefer macroeconomics. And one of the things that I would enjoy it was that game theory, because you really see all the age and the attacks or in terms of you said, it is something that I believe I know. One of the thing that I always say is that if you can impact a nation so and that is, it can impact toward and you can change the entire macroeconomic landscape. So I think that whatever you have to do is to start with the way that the internet is thinking because, you know, that design is rational. So the way that the impact [00:08:00] impact in the market. So I think that is where you are to start. And I really enjoy this class because it was what I want to focus on one agent. And after that, I can see the entire economic landscape.

Harpreet: [00:08:11] Yeah, man. Michael Microeconomics is my favorite as well. I actually picked up a couple of books recently like all of my my econ textbooks, like I still have them all, but they're all sitting at my parent's home in California. But I bought a couple of them and just have them sitting there looking forward to digging through them. One of them is a microeconomics for dummies because I haven't looked at economics type of stuff in a very long time, so I figure that would be a good place to start. And then another one I got was it's called the economics book. I'm looking at it right there, and this one is it's really cool. It makes it really interesting. It's got a lot of lot of pictures and stuff.

Fabrice: [00:08:47] So nice.

Harpreet: [00:08:48] Look, I'm kidding me. So I say, you grew up in Haiti, studied and then Papua New Guinea. How did you end up in Papua New Guinea? And just, you know, for the people like me who are not as good as geography, where is Papua New Guinea? I imagine that somewhere near Australia, but I don't know if that's true, is it?

Fabrice: [00:09:07] So does that come off of Australia? So we used to. I mean, there was a three hours flight to Brisbane, and if you want to go to Sydney like four hours, right? Yeah, it's close to Australia. That's a good question. So just let me just be start with when I was in Haiti. So after high school, I was working with UNICEF, so with the UN and I was working as a monitoring and evaluation assistance. It was like following projects. So we're doing a survey to see if any project they have like. The impact that we were expecting is what happened. So that was after the earthquake that happened in Haiti in 2010. So after that, okay, I just want to do something else. I was young and, you know, it's difficult to do more, and I [00:10:00] found that probably the private sector will be the place where I can really put my knowledge and or whatever I get in school. So I thought, OK, that's where I really want to to work. So I probably once I get the experience after that, I can do something. So one of the thing, even though I left Haiti, so I think that my goal is the same. So the different paths that I went through, they were like, So it's just a way to reach that goal. So going to Papua New Guinea, that was that. So that was another opportunity. So when I started with the private sector, so I went to the telecommunication company and I started with one company that the company called Digicel, they bought them and I worked with Visa and working there.

Fabrice: [00:10:47] It was like I wasn't afraid to take any responsibility, so I was in for. And that's what opened for me. That would do South-East Asia because it was when applied for a position there, and I wanted to take the challenge. Of course, I was like, OK, so in my country, so I know it Data. I was a good data analyst. What can I do? So it was like a way to say that, OK, I'm going to test myself. So it was a challenge and I took it. So I moved to Papua New Guinea in 2015, spent about four years there. And after that? Ok, so that's probably. To make sense, and this is when all the talk about science in humans act as a Data science. So as a data analyst, I was mostly using SQL and Excel. Not that there's so much that I can do so and I start reading about deep learning about a clustering concept. I do segmentation just a basic filter. What if the stores and at the same time, there was not personal decision to move to New York? And I was like, You know what? That's what I needed. [00:12:00] And so coming to New York, it was like, You know, what the fuck is that about science? And here I am.

Harpreet: [00:12:08] That's really cool. You have to work at like four for UNICEF, the U.N. that's that's awesome. And it's like, yeah, making massive impact, massive change. So what was it like working in Papua New Guinea like? First of all, I guess that is for my own curiosity. Like, what's that country like? What's it

Fabrice: [00:12:24] Like there? So one of the things was for me, it was that as in if anyone there. It's not like I have a friend and I went there because I wanted to live as friends, so it was like a really new experience. So let's see if so they are like 15 hours. And so communication to family was really hard, especially sometimes when I'm going to work this, when they're going to sleep or when I'm going to sleep, they have this stuff in their. It was really hard. But in terms of the challenge, I think it was worth it because once you get there, because you are an expat, people will not tell you what to do. So you have to figure out what to do because you are professionals. So you know you come with that expertize from another market, so you start working. So you have to show people, OK, this is my experience and I want them to do. So I felt that it was that you don't have any guidance, you know, because as an expat, they just consider that, you know, stuff. And I think this is really what I needed. And after that, the country was fine. So security was an issue. But once you started to get on, you understand it's like any country in the world. So there's places where you will not go, you know, even in New York, the spaces that I will not go. So it was the same there and I was driving, so accommodation was fine. I felt good, but for me, it was mainly about what will I do with that experience? And it was really [00:14:00] that growing myself professionally and I felt that I did. And when I found out that I was not at this stage, I cannot keep doing the same thing to focus on myself, to focus on my future. And this when I decided the last nine days, the path to go

Harpreet: [00:14:17] Then moved into to New York. Man, that's crazy. So twice he kind of left, you know, left. That's huge changes going on from Haiti to Papua New Guinea to Papua New Guinea to New York. That's that's crazy, man. That's I mean, crazy like the good. The best way, because that's really, really living outside your comfort zone. So while you're in Papua New Guinea for years, did you get to travel anywhere? Did you bounce around this country?

Fabrice: [00:14:43] Yes, I did. I did. I went to Fiji a couple of times. I think I went to fish, maybe three to five times Australia. I went through a lot of time because every time that I was coming to New York or go to Haiti or transit to Australia or spend a night there. And sometimes when we have a long weekend, I'll go to Brisbane or Sydney. I spent some time in Singapore because we had an office in Singapore, so I went from there probably two or three months. Oh, even in Fiji. So that's the thing. That company, they have all the offices in the South Pacific. So we have an office in Fiji, Vanuatu, somewhere in Papua New Guinea and we have an office in Singapore. So working there, the chance to basically visit Southeast Asia is at that time I went to Thailand, I went to Philippines, a couple of Malaysia. Yeah, that's another thing. So if I was in the Caribbean or in the U.S., so those countries are really fast. So you will never plan to go just for a weekend to go to Philippines. But living there was like in what I can just take that flight Philippines and you spend one weekend in Manila and after that, you get back to work on Monday or Tuesday.

Harpreet: [00:15:55] That's awesome. Yeah, actually, my my mom's side of the family is from Fiji, like my mom was [00:16:00] born and raised in Fiji.

Fabrice: [00:16:01] It was that Suva.

Harpreet: [00:16:03] Suva? Yeah, OK. Yeah, yeah, there's a lot of Indian people, a lot of Indian diaspora there in Fiji, and I'm pretty sure the Caribbean to like Caribbean's have a lot of Indian people there as well. That's cool, man. Like, I've always wanted to go Fiji. I've never, never been there.

Fabrice: [00:16:17] So yeah, yeah, I saw the Suva in the when I have a couple of friends in. And this stock could just pull and in Australia as men in Melbourne.

Harpreet: [00:16:31] Yeah, yeah. So men, thanks so much for telling us about a little bit about your background, your life there was really interesting. And so you've got some really cool like writing, man, like, I love the pieces that you've done. You love to unpack some of the wisdom that you got in those posts. And I think one that's really important is just the piece you wrote about tips for staying focused. So I think that's a big one for. I mean, anybody, really. But what can you share some

Fabrice: [00:17:01] Some of those tips with us? Thanks. Thanks for liking my writing. So I'm still trying to improve. I've had that tweet that I can share, and I think that is meant to be it. So I was like, You know what, if everything's happening, you need something to stop fucking you stay focused. And that's why I really wanted to write about it. I think no one is to have a plan. So union, when I say, is just like, think about it, they need to have a plan and you have to write it, so you need to know exactly what you want. So once you have a plan, you know you have a goal, so you know, what's the way to achieve it? So I'm pretty sure that that will help you keep you organized. So you can say that, okay, that's where I want to go. And that's the difference. Way to achieve that. So that's the number one. The second one you need to practice and you need to stay disciplined. So whatever you want to do it. And if you say that I want to be a good driver, yes, you have to plan. But if you don't [00:18:00] take a car and probably do a bit of driving every day, you will never be good at driving. So you need to practice and you start that practice for ten minutes, one day, a second day, five minutes, you need to be consistent. And the last one, what I can say, just like you have to act, you need to make a decision, you need to take action on it. So you have to plan your practice. You want to get better at it. And after that, you cannot just sit up or I know you have to go out there and say, Okay, now I'm taking real actions towards that goal that obviously that you are.

Harpreet: [00:18:37] So when you're making that transition because you said previously your job was a lot of skill out of excel, then breaking into data science, you had to really upskill in like, I believe, use Python if I remember correctly. How did you apply those techniques to when you're learning Python?

Fabrice: [00:18:54] Yeah, I mean Python. So I have to go to I went to the bootcamp when I came to New York, so I started myself, but I must speak without a full time. So it was like the best option for me. And all that, even though if I go to New York, so I still have to adapt so that you know how to go to pull up to school. So I went to that boot camp about five to five months. So one of the things that I had in terms of the plan was that I want to listen to Data find one of the accent that I was to go to the bootcamp. And while in the boot camp, one of the things that I managed to do almost every day was to do code coding challenge. So there's this website that if you coding because I didn't have a new SQL, so but I didn't really have coding bootcamp, you know, so practicing every day, that's something that really have been proven in Python. So I'll go in the morning and go to one of those websites and whatever challenges it, even though probably I spent twenty [00:20:00] to twenty five minutes in most other stuff. But that's just keep me focused on what I wanted to achieve, because that really helped me to say, OK, if I want to be a good data scientist. So I have to be good, at least at Python and which is one of the programing language that most of the company use. Now that, you know, I don't I don't have any Python background, I have to be good at it and practicing almost everything, including coding, coding.

Harpreet: [00:20:29] Yeah, that's good. My repetition driving at home, drilling it in. So you wrote another piece where you're talking about tips for for public speaking, and I think that's something very relevant to Data scientists that think a lot of us aren't really comfortable with public speaking and things like that. So can you share some tips for public speaking and giving talks about Data science?

Fabrice: [00:20:53] So I think that that one year in school, I used to present a lot. So I think that I learned a lot of them and studying law also helped me in terms of hobbies. And after that time, I used to host a radio show. So one of the things that I always tell people, no one is you need to prepare your speech. You probably know what you're going to say, but you need to prepare. So that's number one to take good because once you prepare what happened, you are really confident. So once you are confident, so you will win the audience. So that's number one. The second thing you want people to really listen to you, so you want people to acknowledge why you stay sane. So I find that you have to be as simple as possible. So as the really want to really like a logistical issue, we just want to use this fancy word. But I think that we always have to assume that in the audience, there's a lot of non-technical people. So we have to talk as simple as possible. So we need to probably explain [00:22:00] what we want to explain the technique that we are using. And even though the public is Data scientists debate, we still have to explain the method that we are using. And the last thing is that we just have always to focus on the inside and try to explain it. So once you really know where you go in and you say that, OK, that's what I'm doing, I think that just improves your confidence. And it's just like, make people, people just want to listen to us. And I think that once you see that people want to listen while you are saying, I think that you will be good at it, it will just come because, oh, wow, I'm really interested in something and people are really receptive to what I'm saying. It will just go. And based on my experience, I find that it always happen.

Harpreet: [00:22:46] That's a good point. So if you're going to present on a topic, you need to be excited about it. You need to love the topic yourself because when you're presenting right, if you're up there presenting and you're not interested in it, then people aren't going to fall asleep, right? I like that key point man point. All right. Thanks for sharing that. So. And I want to get into some of your projects because you've done some really cool ones that I've that I've seen that I've seen right about as well, kind of out of the box thinking and creativity for your project. So there's one project that like where you take your econ background and your level of economics and you did it. And essentially, I believe you did it in Python as well. But use Nash Equilibrium using data science and data science methodology. Talk to us about talk to us about that.

Fabrice: [00:23:37] I know that do of that one, I could exact day, so I felt that. So when one of the classes that I really enjoyed was Team two, so you just mentioned it. So when you want to talk about something, you really have to be excited about it. And I think that came to always keep me excited. So when people are talking about, I'm like, Okay, I want to know [00:24:00] more. And I just started with the bootcamp and I to write that article. You what? I want it to be something that I really enjoy. I didn't even know if I would be able to link those two, but I started doing some research on what it is about mathematical models and that time that exactly the same as like, you know, there's a way and a good thing. I found that package. I think that it must be a spy and all that. Ok, so I wanted this job for me. And you know, one of the good things happen is if we a programing language. So there's a lot of library, so sometimes you don't really have to start from scratch. So if someone did their job. So I found that Typekit and I was like, I try to understand it. And so at the end of the day, what I managed to find, just like a write program where you can just give the machine two different options.

Fabrice: [00:24:59] So basically, I mean, I'm not going to detail for the Nash equilibrium, but basically for the prisoner's dilemma. So within a and B. So I just give the two choices and what kind of the ActionScript serve. And I think that that one was really good. But the takeaway of that of that article was as a Data science is the number one is the business understanding. Is that about Python or anything just unique it on the. And the the the business and most of the time, the way that the agents are interacting, you're not going to like to do it. So I think that if you can really find a way to incorporate computer to your data science study, so that will just give you an added value to your background or to your knowledge. So I think that once I did that, I was like, You know what? So [00:26:00] what I'm doing is in my day to day job. So sometimes I just use it to understand the way that people are interacting or the way that the consumer interact into the business.

Harpreet: [00:26:13] That's interesting. I'd love to dig a little bit deeper on that. Actually, just for a record, I interviewed Dr. Kevin Zelman, who we did an entire episode on game theory. Yeah, that should be out at some point in the near future. So definitely check that out, but talk to us about how. So I guess a brief overview of prisoner's dilemma, just like a brief overview of prisoner's dilemma and and like, how have you seen the prisoner's dilemma work out like in the real world with people you're working with, if you have an example for that?

Fabrice: [00:26:46] I mean, just so the prisoner's dilemma is like, so you have the prisoner, so. You want to sell the police caught them and just make it fun, but it's got them and you don't get the police, they don't know. Comment felony. So and this guy gets this idea that, OK, we can't afford to have them into separate homes. And that's what in here. They call in the information so they don't have any information. And the idea is we're going to push them out here if they don't say anything. So basically, two of them, they're like innocent. So they go. So no. So basically, if I if I want to put into the number, so that would be zero, you fuck with them so they don't go to jail. The second one is that if the two of them talk, that's mean that is a reason to be detained and they're going to be say that, isn't it? Is it the thing? So the both of them will go, let's say, with four years of prison. And if one talked, though, they want to keep quiet, so that would be the one we didn't say anything would go with three years of between year because they don't want to see something [00:28:00] about it. So but what happened in the Nash equilibrium is just like this of them because they don't have any information, they are not in contact of any contact. The way that they would think is just like, Oh, you're the guy with something above me.

Fabrice: [00:28:16] So I better talk so I can get out of the way. And I did of the day, the two of them, we just admit that the other one did and they would just go away for probably four years of prison is when if the two of them didn't say anything, they would be like, Okay, you are free. And so when I did that Python script, I entered this number zero zero. I think I say zero three, four four and I get the four for that as a master. I find that in real world, I probably don't see it more often. I've had that. My other, my probably environment is really friendly, but I've seen it in politics. So I sent it a lot in politics like you expect a group to say something and you do something, but overnight because they don't have any information. And then the next day you see seen with some like crazy. Wow, what did they say? And when you think about it, it's just like people didn't have any information. So they would just take the worst decision when they think that probably is the best decision for them, but it's not the best decision for the group. So that's one of the seminars. Equilibrium is just like the economic agent is making the best decision for them to have the best payoff, but it doesn't mean that is the best PR for the entire collective.

Harpreet: [00:29:42] That's all about the incentives, I guess, aligning incentives. All right. And you got some really, really cool projects that that you've done that you did. This one really used machine learning to analyze the Matrix movies. I thought, That's cool. The Matrix is like one of my favorite favorite movies. So [00:30:00] talk to us about this project. Like, how did you get the idea for this project? And I guess what was what was your big takeaway from it?

Fabrice: [00:30:07] Ok, that one is still ongoing, but

Harpreet: [00:30:11] I love

Fabrice: [00:30:12] I love The Matrix, so that's why I my best. I don't know how many times I've seen them. And as we mentioned before. I always like to do things that I like, because it's just keep me motivated. And if I had to present it, I would be like, Yeah, I'm really excited to talk about it. So that was like I probably almost done with the first stage. So I have all my Data ready. So which is like the script for the three movies. So I clean them and there's basic Data Data just like fun number of words for a different purpose. Yeah. And I have some craft, so but what I want to do? There's two more than that. I'm going to use the first one. I'm going to do natural language processing to do a bit of topic modeling. So to see that what is really what they are talking about, I just I know that everybody, they have an idea about the imagery, what I it is. I tell myself that with machine learning, I can really find out what they're talking about. I think that at this stage, I'm going to start in the second one. I'm going to use graph theory to analyze the different positions in The Matrix. So what I'm hoping is to find that Neo will be in the middle and everything is linked to Neo. That's because, you know, that Neo is like the one. So I'm like, Yeah, that's what I want to see. So I just open when I'm two, when I do that graph and I see Neo in the middle and on the edge and that, yeah, that's what I want. If I can see that, I'll be so happy.

Harpreet: [00:31:52] So how did you like? How'd you get the Data for the movie scripts? Did you have, like what?

Fabrice: [00:31:59] Skipping I? What I [00:32:00] use was skipping. So absolutely, I think this is one of the tools that we have as it are science to gather data. So so I went to that website and they have all the scripts, so I get them. Don't think we're cleaning that up to do after we move in the software. It's just, you know, sometimes. So for a speech you love, like, let's say you are Morpheus and you have what they say, but sometimes you have a line to explain that. Ok, so you have to do anything we doesn't have, let's say a name before, so probably up to date. But at the same time, I want I have to keep them because it's about the movie. You know, it's just give some context to the movie. So in terms of topic modeling, so I would need them just to for something I don't know or I will use them now. Yet, I mean, I don't know how we see them now, but I know that I have to keep them. So basically, once I have, like all the speeches and all the different action that happened in the movie,

Harpreet: [00:33:06] That's that's how you do a project. You got to do it based on things that you find interesting, right? That really gets you excited because that's how you get that intrinsic motivation to just continue working through with it. And another one would be really interesting projects. I really enjoyed digging through. This one was applying machine learning to hip hop lyrics, so I thought that was really cool. So talk to us like walk, walk us through how you came up with the idea for this project.

Fabrice: [00:33:34] Oh, so that this project was like my last project at my bootcamp at school. So we have three weeks to do the project, and I wanted mine to be mind blowing. I was like, You know, I'm not doing any this project. And as you just mentioned, so once you do a project about things that you like, it just keep you motivated because I know when you start in the project [00:34:00] as a bit of cleaning and it can be tiring, but if you are really interested about it, so you just keep going. And I'm a hip hop fan, I always agree with fans like Eminem is good and this is better then. So I think that most of the argument about an artist, what kind of music they are subjective. And I wanted to say that was the objective for you to to classify the music or as objective way to analyze the music and the science. So basically, it was like a basic mix of something that I love and did science, and I wanted it to be like art. And I think that's where it came up. But you said that before I did it, it's not something that I always wanted to do. Just like for that project to explore project, you know, I'm going something big and I think it was big.

Harpreet: [00:34:56] So what kind of I guess, what was the what was the problem statement? What methodology did you use? Like, did you did you grab just like the the lyrics or did you grab like the audio? Or did you combine audio and lyrics? Like, how did you how did you kind of piece that project together? Like, what was the big question that you're trying to answer?

Fabrice: [00:35:17] So the big question that I was trying to. Is to to move from a subjective aspect of music classification to a more objective and scientific one. So basically, I was like, OK, there's differences in Hip-Hop, because for someone who is not like whatever song they say, that is the same thing. Let's say that for me, my all Hip-Hop songs are the same, but I was like, You know, that's probably what a lot of people think about it. But what's an objective way to show them? It's not the same. There's a difference in Hip-Hop, and I think that that's that was my main problem statement. I said I can show that they're different in [00:36:00] Hip-Hop music. And the other thing I wanted to show that over the years, if Harp songs are not the same. Those are probably my two hypotheses for this project. So all the is in the Data, it was like a hard one because the main question was when is someone is a Hip-Hop artist? So I found this website to like more than 300 artists, so that would be anyone with at least a pop song. So I started by getting all those names first. And after that, I moved to a genius idea. I think yeah, as it was scraping, was scraping to get those lyrics. So basically artists and lyrics and the lyrics on the song got there. And after that, I went to Spotify and I do an API call to get all the features. So I don't know if you work with any Spotify Data yet, but basically they have all the information about you. So anything. So when I did that, I didn't want to keep this with lyrics and order one photo. So I only keep any artist with audio and lyrics together and something that was really unfortunate because I can say for someone like Jay-Z at that time that any song on Spotify because it isn't possible. So, you know, I don't see it just doesn't matter.

Harpreet: [00:37:26] Yeah, Spotify API, they've got a lot of really, really, really cool like data points about all the songs. So I was really excited when I saw that you did that because that's like a project I always tell students to do is go to Spotify user API and just get all this track information about songs and just do something. So did you have any type of criteria for which songs to include and which song not to include? I guess so.

Fabrice: [00:37:54] One of the criteria that I have it was that the the song has to be in English. I [00:38:00] mean, because of the the package that I was going to use for natural language and it was be too much for me to have increased in French to one. So let's say someone quite sir. It's a yes song in Creole. So I would just keep in English, and I found those with artists as well as Spanish, even if the song featuring so that that's mine. So I have to keep on the English. And after that for my plan is just to make it fair. I took it on the one with more than twenty five songs because I didn't want I didn't have to to give some artists some advantage over the other one because someone has, like 100 songs. They say that is unique. Words will be less than the guy with only 10 songs, so I have to have a benchmark. So I said, OK, you know what? I'll just give any more. And even if a year of a song, I'll just take out the other. And of course, you see, I mentioned Jay-Z was probably my biggest, the biggest missing in the project because I didn't have any songs with Jay-Z.

Harpreet: [00:39:09] Yeah. So that's that's you're making some good selection criteria for what you included or not included in your sets. That's. And it sucks that they couldn't get Jay-Z in there if Tidal had everything he has. I wonder if

Fabrice: [00:39:24] I find that the more that they move now to Spotify yet?

Harpreet: [00:39:30] Ok, yeah, I've checked that out. So what was like the surprising result? What was like the big, big takeaway?

Fabrice: [00:39:37] Yeah. Biggest takeaway was so I have some. I was expecting it, but when I saw the number was the same, so the biggest one was like, there you see. So it's like the ratio of unique words it can against certain words that they are using. So basically, I found out that nearly six of the word used in rap only are unique because [00:40:00] I fell in love with. It was like 300 words versus what it was at 50. And I have to go. I mean, I mentioned about this understanding. So bit of context was why. And after that, I found out that, of course, because, you know, to make the song catchy, so they have to repeat some somewhat. So some songs, there's a lot of well, that they repeat just to make the song. I mean, that's the essence of hip hop. One thing that I found that one I was expecting it was there was a lot of. So that's why in my cleaning, so I have to. There's a lot of wealth like in or so to transform them, because how isn't something and you have like in or two? So I just give this title like b word. So and I think that in terms of the methodology, the cleaning was retiring because asked to create enter into a was for it. So basically somewhere like, Yay, yay! And you know that a lot of the didn't write them properly, so I have to correct them. Yeah, that was that took me a lot of time and not particularly wise act of the year. The topics are evolving because I did the topic modeling, so I found that over the year, the songs are not talking about the same thing. So, for example, we see that starting. It late 20s, I mean, 2000, probably 2004, 2005. The songs were more about sex and drugs, and if you go in the 90s, it was more about street violence and stuff like that. So that one was that gave me a real understanding of it. So that's what that was like. People are the biggest.

Harpreet: [00:41:53] So fascinating. Do you like it? I love these really interesting projects that you work on. So like, can you share [00:42:00] some tips for the audience here on how to come up with project ideas? And you know, like what? How do you come up with these ideas? Like what can what can the audience take take away from from from this so they can go create some creative stuff for themselves?

Fabrice: [00:42:17] I think that coding and math, the entire machine learning models today are really complex. So pick something to have fun if you don't want to be stuck with coding enough to start working on something that you don't like. So you need to take a project that will have fun you just enjoy. And the second thing is, don't be scared of the challenge. And I think that there's enough tools and techniques in machine learning to address a lot of issues. So I don't think that anything is

Harpreet: [00:42:52] To

Fabrice: [00:42:53] Too small or I cannot do that. Try it. Don't be scared of the challenge. You have whatever idea that you have. Go with it. Worst case that can happen. You will need more data. So you just have to get more data. Or while you're doing the analysis, you find some issue. I mean, at least you know about it. So don't be scared of the challenges so that we have enough tools of science to tackle a lot of problems. So go for it.

Harpreet: [00:43:21] I love it, man. It's the right mentality, I think people just they. The two afraid, too afraid to go, pick up a project and just do it right, you just have to be limited by your own creativity, I think when it comes to to do projects, right? All right. So you've got definitely unique projects, but unique projects have a unique set of challenges, and that's all about collecting data. Right. So well, can you share some tips with the audience to help us make sure that we're doing a good job collecting data? Do you have some favorite places that you go, some websites or anything like that?

Fabrice: [00:44:00] I [00:44:00] found that once you have a project and you know what you are doing, so you have your hypothesis that you need to identify your sources. You need to know that these are the trials where you're going to get your data. Don't even start with collecting data, so just identify sources just as give you an idea of what is available so you know what you can get because with web scraping or using selenium or even API. So there's that opportunity to get data so it can just get back to the challenge. Don't be scared to take a challenge. Just go see that. Ok, it does. It's fantastic. I'm going to try it. And another thing is never assume that your data collection is final. Sometimes you will call the project and you'll be like, OK, I get my data. That's it, and you just want to start coding and do your machine learning model or anything. That's all. You have to spend a lot of time doing the analysis because sometimes the analysis just tell you that you need more data, so you have to be ready to collect more data. And I think that's something which is just really interesting.

Fabrice: [00:45:15] You just have some idea about simple to understand because sometimes you probably want to have data about the population, so you might need to take a sample and just that give you you can do extrapolation just to get an idea of the population. So it's really good to understand or sampling work. And the last one that I see, which is that we make the entire data collection good. You have to make sure that your data is good quality. Never was to just go to the analysis or to start with your linear regression or whatever you want to do or I'm going to do random frauds. Ok, let's do it. No, just make sure that you have good data. So [00:46:00] I think that that one is if you don't have a good data, so you don't you cannot expect your model to work better. So once you realize, oh, I have like low accuracy. Yeah, but what about your data? So I think that good quality of data is, I mean, that's that's key. That's the key. If you don't have good data, nothing is worked. So that's one you have to be focused on it.

Harpreet: [00:46:24] Right on, man, it's some great tips. I mean, you've got some really, really, really cool projects that you're working on. I'll be sure to include a link to your GitHub in the show notes, so that people can check that out. But I like this company that you're working at. Way better. Like this goal sharing network? Like, that's pretty interesting type of stuff that that you have to do in there. How do you guys use data science to to to help people like meet their goals? Like how does that work

Fabrice: [00:46:52] At this stage? So we are a startup, so at this stage, we're not really using data science to help people, but as a data science working there. So in terms of the decision that we are doing, the number one is just using people data, so basically just a backbone of the alphabet. So we are helping people stay motivated with games. So let's say that every Monday or every week we have a new set of games. So in order to propose a game, we need to know what the game that people love. So we are using the data collection or data analysis and data science to understand people's behavior, to understand what people want to, what people want or what will motivate people or what are the most successful games. So if a game of every week, only one player in it. So there's no need to use it. So I think that at this stage, we are using the best science tool for the selection of things that we are and as well. So because it's is so desolate [00:48:00] of social endeavor, so we are not analyzing the social feed to understand what people are talking about. And I think that that's something that we're going to use every day in the future. But we have a lot of data and I think we want to make people better at something. And the only way to do it is by knowing understanding people behavior. And what they are doing when they get to the absolute value proposition to give them.

Harpreet: [00:48:32] Sounds fascinating, man. I got to check that out. Like gamifying. Oh yeah, I think that's really cool. Yeah, we'll do a last question before the random round here. So it is one hundred years in the future. What do you want to be remembered for that?

Fabrice: [00:48:50] That's really window. I said that. Yeah, why not should? I said that I want to be remembered for someone who loves the people running and wanted the best for them. So that's something I would like. Yeah. Just guy always wanted the best for people, and because of talking about it last time I said that I just want to have some impact in the way that they are teaching the science. And I'm working now with this organization where we're just trying to get that science in Haiti. Yeah, that's something that I really would like to be big and maybe 900 years people remember me for that.

Harpreet: [00:49:29] That's awesome, man. I dig that. So let's listen to a random round here. First question is, if you were to write a fiction novel, what would it be about and what would you title it?

Fabrice: [00:49:42] Yeah, I love love what I bought is a secret agent. Yeah, so I think that it will be some secret agent. Maybe it wasn't me first lesson from that, and it just get back to the were [00:50:00] too. I don't know. Yeah, that would be about agents. Not not not agents, one type, but more like a mix of a mix of James Bond and Rambo. Maybe.

Harpreet: [00:50:12] Yeah, that's a good one. That's a book I would read for sure. So when do you think the first video to hit one trillion views on YouTube will happen and what will it be about?

Fabrice: [00:50:31] I don't think it would happen so soon. We could. Yeah, I mean, that's why I say that. I mean, just doing the coffee time. So then like, look at for what's five seven months and no video is that that shows and it was like the right amount of time for people. That's exactly what people are doing, Netflix and YouTube. So I think that it has to be something like maybe some some people with like maybe a lot of followers on TikTok and they just come with that new thing and it will be just happen. I don't think that any of this video with like, be done through it first. I don't think I don't think that Despacito it.

Harpreet: [00:51:17] Yeah. Esposito has about seven point zero. Four billion. Yeah. Shark Baby Shark has seven point zero five billion.

Fabrice: [00:51:31] What that does with with that with Wiz Khalifa and Charlie Puth,

Harpreet: [00:51:36] See you again. See again. Yeah, that's four point seventy nine billion. Far, far away from trillion.

Fabrice: [00:51:45] Yeah, I think that yeah, no doubt it'll be a new one.

Harpreet: [00:51:48] So what are you currently reading?

Fabrice: [00:51:51] Oh, now I'm waiting. That's it. The book is Artificial Intelligence in Practice by Bernama Silicone. So basically, [00:52:00] what he's doing is taking all those big company like big companies in this so and so, or they're using artificial entities as data science

Harpreet: [00:52:10] To

Fabrice: [00:52:11] Increase their production or to increase their welfare. Yeah, I like those because it's just really helped me understand why what I can do data science is really practical. It's not about the killer thing, it's really practical. So I really like it. It's a good one.

Harpreet: [00:52:29] I'll take that one. What song do you have on repeat? I started this

Fabrice: [00:52:34] This idea that listen to any song. I'm just trying to catch up with a couple of podcasts, so I think this book was to start a podcast. Yeah.

Harpreet: [00:52:46] How about this? What song is like always playing in your head?

Fabrice: [00:52:51] Well, because it just said, if you suck, I think that

Harpreet: [00:52:54] It's been, you

Fabrice: [00:52:58] Know, I've heard that the last album that I listen is the Eminem album. Ok? Yeah. And I said, Yeah, yeah, yeah, music to be murdered by.

Harpreet: [00:53:09] So that I would think that would have to listen to yet. I've been on some weird music lately. I've been on some deep instrumental house stuff. Yeah. So we're going to do the random question generator. Oh, wow. What issue will you always speak your mind about

Fabrice: [00:53:29] What is going on this week?

Harpreet: [00:53:33] Pool? What is your theme song, right?

Fabrice: [00:53:38] Yeah, I felt that that would be Eminem. I'm not afraid. Yeah, that's a good one. But that one is easy.

Harpreet: [00:53:46] That's a good one. What are you interested in that most people haven't heard of?

Fabrice: [00:53:51] I think that people might be free diving, free diving.

Harpreet: [00:53:56] Yeah. Oh, OK.

Fabrice: [00:54:00] So [00:54:00] basically, you've got that thing with any. I'd probably read about scuba diving. So scuba diving, you go underwater, but you have all this equipment, but free diving, you go without anything. So one breath and you just go down him.

Harpreet: [00:54:15] Do you do that quite often?

Fabrice: [00:54:16] Yeah. I mean, last year I didn't because of COVID. But yeah, the last time that it was now. Yeah. Yeah. So basically, you're just at the surface. You you do a couple of breathing exercises, and after that you take that deep breath and you just go down. I mean, some people do what we do, but that's really fun.

Harpreet: [00:54:41] Wow. How long were you able to hold your breath for? Okay.

Fabrice: [00:54:46] The last time I went, I was probably two minutes between two and three. But when I started, when I started, what, three years ago or four years ago? Yeah, it was only one minute. My max was one. But the thing is, what if you still for diving and you're still doing exercise, so you just getting better at it? But everything is just about knowing yourself and staying calm and try not to manage a CO2 in your body.

Harpreet: [00:55:19] Wow, that's fascinating, man.

Fabrice: [00:55:21] Yeah, that's cool.

Harpreet: [00:55:23] What would you do on a free afternoon in the middle of the week

Fabrice: [00:55:28] Or reading nice? Yeah.

Harpreet: [00:55:32] What's the best thing you got from one of your parents?

Fabrice: [00:55:38] Of course, they got fired up. I think that I like joking and I got say that I got it from my dad, always make a joke. Yeah, I'm really annoying.

Harpreet: [00:55:51] Press on the last one here. What incredibly strong opinion do you have that is completely unimportant [00:56:00] in the grand scheme of things,

Fabrice: [00:56:02] Something that I. Yeah. So one, yeah, that's really fun. I find that you don't have to have an opinion about everything. I do believe that. And if so, that's something that people are talking about any subject. And if I don't know anything about it, yeah, because I something that, yeah, that's a lie. So you don't have to have something to say about anything. That's a really strong opinion that I have. And I mean, whatever happens, that's not going to change.

Harpreet: [00:56:35] That's an interesting, strong opinion. You yes, a strong opinion on anything. Actually, let's do it. Let's do one more here. Pancakes or waffles?

Fabrice: [00:56:47] None of them. That's what I don't like. I'm not American,

Harpreet: [00:56:55] So I guess that makes sense. All right. So that was it was awesome. So how can people connect with you and where can they find you online?

Fabrice: [00:57:06] Oh, so I'm on Twitter and I.T., so everything just my first name and my last name. Same for my midterm, in my midterm. So now I took a post post about an article because I'm preparing a couple of new articles about the science and I'm going to read them soon. So my means I'm saying Fabrice Mazdoor, my LinkedIn is the same. If you could just go to my link and you just find the link for my online presence.

Harpreet: [00:57:35] Definitely, Matt, I will be sure to link to the all your various places in the show notes as well so that people can connect with them directly for me. Thank you so much for taking time out of your schedule to come on the show today, man. I really, really appreciate having you here.

Fabrice: [00:57:49] I have to thank you. And I mean, that's a good job that you are doing. I really appreciate it. Keep doing it, man.

Harpreet: [00:57:56] That's my pleasure, man.