144: Why Should You Build Projects as a Data Analyst (Thu Vu’s Story)

None of the projects that I posted on my channel, I knew beforehand that it would work. It was just sometimes it's completely absurd. And I thought, yeah, like, how could I make it work? And then several days, like tinkering with my code and try to like, look at other tutorials, look up things on Stack Overflow and see if anyone has any. ever done something like this. Yeah. So it's also a lot of like findings for me. Sometimes you have to be creative and solve your own challenge and your own problems because yeah, you always encounter something in your project. The good mindset is just, uh, like there's got to be a solution. So don't give up. When you first see an error or see like a problem.

All right. If you are watching on YouTube or you've ever looked up data analytics on YouTube, you've probably seen, uh, Tu Vu, our guest today, one of her videos, because they are absolutely amazing. Uh, in the past, she's been a data analytics consultant with companies like PWC, uh, and she's a prolific. Content creator in the data analytics space to welcome to the podcast.

Thanks, Abby. It's, it's really, really a great introduction. So kind of you. Of course,

it's

my pleasure

to be here. Of course. I'm so glad to have you here. One thing that I love about your videos, uh, is you do some pretty cool projects. Uh, you do some pretty cool. Things with, with ai, things with machine learning, things with just data analytics and data science in general. And I think we need more of that on YouTube, more like actual projects being done. I think you do a great job of doing that.

Yeah, absolutely. I think, um, people talk about data science or machine learning or AI a lot, but I think what I personally missed, it was like some kind of like a hands-on demonstration of how you're gonna use a new technology, let's say, uh, like a, a network. analytics or some AI, some cool AI models or large language models, how you can apply it to your own problem and also demonstrate it like end to end almost, um, how you start with an idea, how you get, um, inspired, how, uh, how you think about the problem, how you frame it and how you Kind of go step by step, explore it further and further. And then in the end you have something that you can show to other people. And hopefully it's a little bit useful and um, yeah, hopefully you have fun. So that is kind of the idea that I kind of like, yeah, it was not, The first thing that I, um, actually started when I, um, yeah, when I started making videos on YouTube, um, it just occurred to me that people really liked it and people really appreciate the effort to think something so much, uh, yeah, think through, uh, some, a particular topic in such a great detail. Also kind of hopefully inspire other use cases for people to try out. Yeah, so that, that was kind of like the motivation. And up until now, that's kind of like, that is the type of video that I really like making, although it takes a lot of effort. And of course, like I lost some hair because of it, but yeah, it was fun and hopefully helpful for other people as well.

I think so. And I think it's very impressive because it's not easy to make technical videos and make technical videos engaging for like a YouTube audience, which, uh, I think a lot of people watch a video for like 35 seconds and then skip to the next one. So your, your ability to, to kind of capture these technical things in a, you know, shorter video, but not like too short, uh, I think is very impressive. Uh, have you always enjoyed making fun projects like this?

I think at the beginning, it was kind of a struggle. I think it was like, yeah, I think without a lot of, like, experience with, uh, you know, like making tutorials, it was also kind of like always like climbing such a, like a big mountain every time. Like how you talk through everything, how you explain everything, every little step that you make, uh, making the, like the screen recording and also kind of like, So kind of like nice B rolls, you know, like in YouTube, you have like all these kind of like fun thing that you film yourself and then combine it in, in like a good storyline. I think that that was kind of a struggle for me in data science or machine learning. You can always find some kind of like a project in terms of a blog post or, um, like a Jupyter notebook or a GitHub repository. And that, um, yeah, that's usually how people think about these projects. But like, how you present it in a video in an engaging way, I think it was like the, the biggest, yeah, it was a challenge. At the beginning I was scared. I was totally like, I, I didn't really, I could not really enjoy filming myself and like make such a complex explanation. I was kind of uncomfortable with it. At the start, but it got better and better. I know how to kind of like prepare my scripts, how to kind of like, think about it, maybe a few days, come up, come back to the project and refine what I want to tell. And, uh, it, the workflow gets better. Better and better. Uh, so it's also less scary and I'm also not a native English speaker. So sometimes there are a lot of concepts I want to explain, uh, and I like words for it. I love the way to, to explain it. And that's also frustrated. Uh, but yeah, it's, uh, yeah, all of these struggles, I think it's kind of like, it's worth the effort for me and I think it's really rewarding to, to create. That's kind of my, uh, like how, how it went for me.

It's, it's really impressive, especially, especially like in a, in a second or, or how many languages do you speak?

I speak, uh, Vietnamese as a, uh, as a mother tongue and my, uh, yeah, definitely the next language is English. And I also speak Dutch because I've been living in the Netherlands for the last 10 years. That's so

impressive. That's so impressive to be creating this good of content in your second language. Now let's talk about a little bit more about projects. So one of the cool things that you've done is you've built a project to analyze your finances, uh, with like a local LLMs. Like you basically created your, your own version of like chat GPT to specifically look at your finances. Now that's crazy because. Like I think most people would probably just like do something easier to do that, but you're like, no, I want to make it cool. I want to make it hard. Have you always been interested in like creating kind of cool personal pet projects like this? Yeah. Yeah.

Yeah. Thanks for pointing it out. I think that's also kind of like one of the projects that I got some really, like some people saying on YouTube. Oh, like you, you just invented something that was completely unnecessary because like on the, like the bank banking app that you are using, probably you also have kind of the insights feature where you can also have the same, do the same thing. But yeah, yeah, I know about that feature, but I was like, huh, how can I use, uh, an LLM to help me with this? And because I download my bank statements all the time, uh, and I like in. Yeah, I just like looking at it myself and see in detail what kind of expenses I make. Yeah, so that was the start of the challenge. And, uh, yeah, that video got a lot of, uh, nice, um, uh, yeah, nice feedback because I think people really like to have feedback. Something that is a bit private, like, uh, when you have an LLM and you, uh, you cannot post your blank statements on ChatGBT or Cloud AI to help you analyze it. So yeah, like using a local LLM is a great way to kind of like test out this idea and see how well it works. it might work. And I also have, yeah, it was like a, like a trial and error. I also didn't know if it would work, um, at the beginning. I, I think, yeah, none of the projects that I posted on my channel, I knew beforehand that it would work. It was just sometimes it's completely absurd. And I thought, yeah, like, how could I make it work? Uh, and then several days, like tinkering with my code and try to like, look at other tutorials. tutorials, uh, look up, uh, things on Stack Overflow and see if anyone has ever done something like this. Yeah. So it's also a lot of like findings for me. Sometimes you have to be creative and solve your own challenge and your own problems. Um, because yeah, you always encounter something, uh, in your project and, um, the good mindset is just, uh, like there's got to be. And a solution. So don't give up when you first see an error or see like a problem. And that project definitely didn't work well at the beginning because I know like the LLM was was really unreliable. So I had to like change the temperature for the. LLM, and then tweak something in the workflow and try to validate the output of the LLM with Pydantic and all these kind of things, uh, just for a toy project. So I was like, yeah, it was a lot of work. Um, but it was, it was fun. Um, yeah. Yeah, I think nowadays there are so many new frameworks that help you do these kind of projects, maybe like in an easier way, or like, um, Python packages that you can use, I think, like, instruct, um, like some new Python packages that lets you output things from LLM in a structured format. That is something that I only knew later, but that was much later after I posted that project.

Yeah, I think there's so many good things that people can take from, from what you just said, because I think oftentimes people, you know, look, look at you and maybe look at me and they're like, Oh, these people are experts with data. They know what they're doing. And the truth is that no one actually really knows a hundred percent what they're doing in data ever because it's ever evolving. It's ever expanding. Uh, there's always new things. And. Uh, you know, I, I can't, I mean, I guess I can speak for you cause you just mentioned this, but like, uh, I get stuck all the time and it's still like a process to, to troubleshoot and to, like you said, use chat GPT to try to solve or go to stack overflow and, and get through those problems. So even though you're creating these videos, you've done a lot of them, you have like, like a decade of experience, you're still getting stuck and you still have to troubleshoot.

Yeah, exactly. No, I, I think anyone can do this kind of projects. Um, yeah, given that you put in the time and put a little bit effort and some patience. Um, and I've seen also a really, really cool project on your, on your channel as well. And I, I thought like you really put a lot of attention to all these details on the videos or visualizations that you've made. I thought like, yeah, it's really. Really cool. Yeah. Maybe something sometimes I just thought, okay, I can create this visualization of like, um, you know, animating something just, just for the fun of it. Uh, and you did it sometimes. And I thought, Oh, like you have some really, uh, really great, uh, insights on how you can show things differently. And so I think, yeah, like, yeah, with making YouTube, it's, it's really fun to look at other people's, um, work and see how you can learn from them. And yeah. And I guess for many people as well, uh, in the audience, yeah, you can definitely just like sometimes come across something on YouTube and then you thought, huh, I can maybe do this as well. Um, yeah, so that's a great way to learn from each other as well. And it's definitely, I'm not a, yeah, know it all kind of person, definitely in data science or

machine learning. Well, you definitely know a lot and people can learn from you, uh, a lot and, and yeah, I've made some project videos. In the past. I haven't made any project videos recently. I find them that it appears that the YouTube audience doesn't like them as much. I know I've, I've spoken to Luke Bruce in the past as well. And he has like this awesome video on his channel where he was analyzing his mountain bike data. And I was like, that video rocks. And he's like, I know, right. It should have way more views. And I don't know. It's, it's, it, I think you did a great job of, of making it digestible for the YouTube audience. Cause I think these personal projects where Like, for example, I, I looked at like what states Google the most every single hour for like a quarter of a year, you know, you've done this analyzing my financial data with an LLM. Luke's done the mountain biking one. I think these projects are fun. And I think, I think it's as data scientists or data analysts, like we want to use data in our real normal life, not just in our work or our business. So these types of like personal projects, I think can be really fun.

Yeah, yeah, absolutely. Um, no, I think, I think you're right that, uh, not all the videos that you make or all the work that you make would, uh, get the recognition that you think it deserves. Yeah, it's, um, it's, it's hard and, uh, I'm sure some people also post things on LinkedIn. I also have some friends who, uh, post, um, try to post on LinkedIn more often, but really like doesn't get much views or, uh, like interaction. And then. Yeah. You feel discouraged. Um, and I'm sure a lot of people also relate to that. I also don't know how some videos got, uh, seen and some people, some videos just got completely tanked and no one really look at it ever. It's, it's really hard to, to, to kind of like predict that even though, yeah, I really want to predict it, but yeah, yeah. I think there's some kind of secrets. I tried to make the first, um, like the opening of the video. really engaging. Like the third, the first 30 seconds or so, uh, that's what I learned from all the YouTube gurus and, uh, try to kind of like make the best edits out of it, uh, and see if people keep watching longer. And usually they do, but overall the quality of the whole project is the video is, is, is more important, I guess. And I hope that is,

that's true. It's, it's definitely hard. Um, let's talk about, so like with these cool projects that we, that we've been talking about, if people want to build their own, obviously they can kind of look at the stuff, you know, you and I have done on YouTube. Um, I know that you give you your GitHub, uh, in the description a lot of time, uh, for most of my projects, you can get all of the GitHub stuff for free. It's kind of like open source like that. But you also just recently created something called Python for AI projects, which basically. is an opportunity, a platform where people can, you know, build some pretty cool projects for an AI with Python, kind of with your guidance. Is that right?

Yeah. Yeah, that's right. Um, it's kind of like my, uh, really my, I put my heart into this project because I believe that many people struggle to learn the basics of Python. Python and AI, because they don't have the really like the most beginner friendly kind of, uh, guidance or kind of like a road map. Um, so that's why I decided to create this, uh, this giant kind of like, uh, curriculum that teach people Python from scratch, all the fundamentals, and also hands on stuff on how to learn, how to, how to use Visual Studio Code, how to use AI assistance for your work, and also learn the basics of machine learning, deep learning, and AI, and with some project walkthroughs as well for people to really follow and create their own projects with kind of like the idea. Yeah, using the, uh, large language models and kind of like, uh, yeah, just like some projects I did also on my YouTube channel, um, one projects about extracting information, uh, from PDFs using, uh, large language models and how you structure it in a nice format in a nice, uh, table format. And also, yeah, I'm also planning to add a few more advanced projects like fine tuning and LLM and all these things that, yeah, I also kind of like, I really always wanted to do it also myself. Now it's also an opportunity to kind of like explore it further and help other people to learn them as well.

Very cool. I, I might have to check that out because yeah, I definitely, I don't know how to do anything with like an LLM from scratch. So there's some, there's probably some things in there, uh, that I could learn. So that would be a lot of fun. We'll have the, uh, the link in the show notes down below.

Yeah, definitely. Yeah. No, I, I think, uh, well, you, you, you know, Python and you, yeah, probably you can pick it up very quickly and, uh, all the machine learning AI stuff. Uh, I teach someone as really like someone who has never worked, never built a machine learning model before. Try to teach the, like the fundamentals and the building blocks for you probably is, is much less relevant. Um, but yeah. That's, uh, indeed. Yeah. That's kind of my project

at the moment. Very cool. I can tell that you're, you're really excited and passionate about it, uh, which I think is very cool. Um, we've talked a lot about projects and, and your, your great YouTube channel, and I've kind of given a little bit of your background, but, uh, I'm guessing a lot of people listening don't a hundred percent know. Uh, your, your background. So could you just tell us like what you studied in school and then maybe what your first job was out of school?

Uh, yeah, yeah. Thanks for, for asking about this. I think, uh, yeah, I, I also haven't really shared about it a lot on my channel. Um, my, about my background. So if

you don't want to talk about it, we don't have to just so you know.

Oh no, no, no. Of course. Uh, no, of course I can talk about it. Yeah,

so don't need

to add it in an edit. Um, yeah, so I, uh, yeah, so my first, uh, degree that, um, at school is, uh, economics. And, um, back then I was, yeah, I was still living in Vietnam. Um, and I got my bachelor in economics and then I go to, I went to the Netherlands to study a master, uh, in economics. as well. So this is, yeah, it was kind of like a very, uh, theoretical degree. And, uh, although, yeah, you learn some basic stuff, econometrics, um, linear regression, which were like basic statistics. And that was quite useful later on as a, like a, when I would start working as a data analyst and then learning a bit more data science y stuff, machine learning. And that was in 2015. So I moved to the Netherlands when I was Yeah, around, yeah, 22 ish, um, back then. I started working in the Netherlands and stayed, um, decided to stay, uh, even though it was really not the first plan, uh, when I moved to the Netherlands to study. Um, but, yeah. I found out that it was like really, really a great country. And I fell in love with the culture, the food, not so much the weather, not so much, but people were great. The working environment was really, really transparent, really nice, very efficient. And also people are very direct. And I really liked the way that, you know, people are honest to each other and, um, and, uh, Yeah, you, they give you really straightforward feedback that you can improve on. When I start my internship, uh, after right after my master's, I felt really, I felt really good about, uh, working here. Um, so that's how I started my career actually. And I started working as a data, kind of like a research assistant in my internship. And then later I got a job offer as a data analyst. So I got really lucky to, you know, Actually start in this career because at the beginning when I learned economics, the thing that I would think about was more like a researcher or maybe working in policy, working in maybe a little bit like even started a PhD and that I even applied for a PhD in Amsterdam. I still remember. And thank God I didn't, I didn't got it. I didn't get it. uh, otherwise I would be like, I don't know where I would be right now. And yeah. And a few, yeah, a few years later I start working at BWC, um, Pricewaterhouse Coopers and I start working as a, a consultant for six months. Six years. There, I also learned a lot of different, uh, new things and worked for different, in different projects and I found it really incredibly, um, really, uh, I learned so much. Um, it was really helpful to work with so many different people and you pick up new things every time. Yeah, and that when I was working there at BWC, I decided to learn a bachelor. degree in computer science, I decided to take it because I feel like I still missed something. I, my technical skills were still kind of like not so, I was not so confident. I was, I was still like, yeah, I was probably, you know, like an imposter feeling and also the drive to learn more in a more kind of like structured way. Um, that's how I decided to take the degree. It's an online degree that you can take via Coursera, actually. It's very nice. Yeah, it's a, it was a lot of work, actually. It was a master, um, bachelor degree with, I don't know, 22 modules. And yeah, I still have the final project that I have to finish. Um, so it was in the end, it was like six years now that I haven't finished. So I still feel ashamed when I talk about it, but yeah, uh, in the end, yeah, I work on, uh, the YouTube channel a lot and, uh, it was all kind of like all go into each other, uh, kind of, yeah. So that's kind of like my, my, my, work and my personal history and like my, uh, yeah, my story so far.

Can you just, can you just submit a URL to, of your YouTube channel to the degree and just be like, here's my final project.

Yeah. Yeah. Yeah. Like for like my own project or

just like your whole, your whole YouTube channel. I feel like that should count as your final project. I feel like they should, they should, uh, give you, give you credit for that because you've done some, some pretty cool things on there.

Yeah, that's a great idea. I will, I will try it out.

That's, that's great. Yeah, I think it's, I think it's one thing that's, uh, I want to just pull from your, your story there, uh, was you going back to school once you had a data job. And one of the things I try to, I, I try to help people who are like brand new to data and who like want to become a data analyst. And obviously going to back back to school is always an option. Um, but a lot of the times if you get your foot in the door first with any sort of data job at the beginning, it's going to be so much easier to go back to school for a variety of reasons. And so like a lot of the cool things that you do, like, like the LLM stuff, uh, use Docker in that video. A lot of that stuff is, is things that you don't necessarily need when you first land your data job, but, but they can help you become a better data analyst down the road. And so I kind of like how you, you kind of gotten your foot in the data door with, with the data stuff you had from your economics degree, and then you, you upscaled after you were already there. So that way you can, you can become a data analyst or sorry, become a better data analyst, you know, have a bigger impact at your company, uh, hopefully get, get compensated more and better because of it. Um, but I love that you did that. After you get you started, basically,

yeah, yeah, definitely. And, and I think this is, uh, you're totally right. It's so much easier when you get an internship or you get a, like a really beginner, uh, like an entry level job in data science or data analysis. Even like us, just a small, uh, portion of your job is, uh, data related. You can always like show it a little bit more that you have some experience. And this is really a big advantage. So, yeah, I would always advise anyone to, when they start, just think, uh, step by step and, uh, take anything that you may find, like, you can learn something, um, regarding the data skills, and then you can go, uh, can move on from there. That's so much easier, indeed.

One of, one of your latest videos, you explored data trends, um, and you found some pretty interesting things, uh, that was going on with the data job market, the tech market in general, what was like your favorite trend that you kind of discovered in this video?

Yeah. Yeah. I think the favorite trend for me is like. The new development, when you think about like technical skills, I find that like Python is, has been really so become so much more ubiquitous. So, so much more universal compared to a few years ago. I think definitely a few years ago, it was like, uh, for the discord analysis or some particular software, like SAS, even if you ever, uh, even ever used to use it. But right now, also within my work, a lot of, in a lot of projects, we are migrating all the code base from SAS, from R to Python. So it was like a nice. An interesting observation, and especially with the development of AI right now, Python is supporting a lot of cool tools. For example, like, um, uh, things like lang chain and all these different frameworks to create, um, an AI powered application, an AI agent, all these frameworks are all in Python. Yeah, that, that, that's really like a Uh, yeah, like a really cool thing to, to, to, um, to recognize. Further, I think there's also some interesting trends that I noticed, um, in kind of like the freelancing space. It seems, it seems like, They're more freelancing jobs than right now, than, than a few years ago. And I'm not sure why, but I feel like companies are more like, probably they are experimenting with things a lot. And that's why you see. Probably some of them have a little bit budget, uh, or even individuals or small business owners, they have a little bit budget and they want to hire someone to do something for them. I recently have a friend who worked a lot on, um, uh, who knows a lot on RAC, so, um, uh, retrieval, uh, augmented generation kind of projects using LLMs. And then, um, yeah, so that person connects it to me. Uh, asking, like, do you have someone or you can help me with, uh, uh, building and, uh, kind of like a tool to extract this and that information from like, uh, a hundred PDFs that he has. And so, yeah, so I introduced my friend to, uh, to that, um, to that person to, uh, to help him with, with this task. And I think this is also kind of like an example of like how people are recognizing the role of, uh, AI and automation. And they want to get some, something done. And so, yeah, it doesn't need to be a fixed contract, a fixed job. It's more like a experiment sometimes. And so, yeah, it's a, I find it also really interesting and I keep thinking about how, uh, how people can find these kind of projects. Uh, they can, yeah, like people who need to get things done and people who has the skill, how can they, uh, meet each other more often or how they can more effectively, uh, meet it, uh, like kind of like, um, come across each other's, uh, and connect to each other. Yeah, that there are the two trends that I, yeah, that I really like. And I also kind of like got a bit surprised, but also not so surprised, uh, how, how that, how, yeah.

It's fascinating. We live in a really exciting time where you can start a side hustle or start your own business. That's, that's what I did three years ago, three and a half years ago was I started to freelance and I started to make more money freelancing than I did in my regular job. And I was like, okay, I'm just going to do that. Oh, really? Yeah. Yeah. That's why I left Exxon was to start doing freelance projects and start an agency. Yeah. I ended up switching mostly to teaching because I figured out I really enjoy teaching. So that's what I do. Full time now pretty much. Um, but yeah, the freelancing stuff is super fascinating and I think there's a great opportunity, uh, for people to get into that. Uh, and I also love that you, you brought up the, the Python trend. I think it just became, you had mentioned that video. It just became the most common or most frequently used language on GitHub, uh, which was a big deal. Um, so Python, definitely a thing of the future. Another trend, uh, that I really liked, especially since I helped people land their first day at a job is you looked at like. The number of data jobs over the last few years. And if like, we've seen a lot of layoffs or if we've seen a decrease in jobs, because you know, a lot of people are like, Oh, the economy kind of stinks. And you know, the job market's really bad right now. Uh, and your conclusion was, you know, maybe it's not as bad as people might say. The, the, the graph was kind of a little bit downward in terms of like number of jobs, but it was relatively flat. And that's actually, I did. Yeah. Uh, a similar video recently where I looked at, um, the growth of, of data jobs from a different data source and the data source you used and basically came to the same conclusion that the, if you compared it to 2019, uh, data job openings were up specifically for like data analysts were still up around like 20%. But it was year over year, but it was a flat 20 percent for like the last year or two. So I was really comforted to see, like, you kind of came to a similar conclusion that I did with a totally different, uh, data set, completely independent of each other.

Oh, that's really cool. Um, that that's really cool to see. Indeed. Um, when I was using, yeah, I actually use the, uh, kind of like the data from, um, from, from, uh, collected by Luke, uh, Luke Burrus. And, uh, yeah, he's the man behind all this, like, web scraping stuff. And, uh, I also, I was also a bit doubting, uh, I didn't want to make a conclusion that, oh, this is like decreasing that we are seeing. Indeed, it's more like flattened out and, uh, depending on how you see it. Um, and as you say, it's more like a, uh, glass half full or empty. You, yeah, like it's quite, I think it's quite normal to see some fluctuation over the year over year. And, uh, it's, uh, it's definitely, yeah, I don't think it's something that I would worry about, but more like, uh, what kind of jobs are being posted? Like, the job compositions are changing rather than the number of jobs. I think. Probably within the same job title, you probably have something new in the job descriptions. And I, I didn't, um, really have the chance to really dive into that in, in that, um, data job trend video. But I think it would be really cool to see how the, uh, the job functions or the job, uh, description is changing. And how you can maybe learn from that. What can you prepare to meet that demand in the future? I'm sure there will be more like, uh, really things that are more like, uh, data AI engineering kind of role that are emerging in data science, in the, like, data science, uh, Uh, machine learning space. And so, yeah, I, I think, uh, yeah, it's probably like, it's better to, to, to, um, a little bit, put a little bit like, uh, yeah, yeah. Take that with a little bit grain of salt. When you look at the chart, um, probably it doesn't really tell the full story.

I agree. It's, it's, it would be really interesting. That data sets very rich. Um, but once you get into text analysis and NLP, you just have to have more data science skills. It's like a whole separate. Part of data science, which just takes longer to do than things like counting and line charts and, uh, bar charts and stuff like that. Um, right, right. Definitely. Which, which maybe it's a, it's a great project, uh, for your Python for AI projects, uh, group that you're doing with, with the course. So maybe that we'll look forward to seeing that. On the curriculum in the future to thank you so much for being on the podcast. If you guys haven't checked out her channel, please go do so. Now we'll have a link to it in the show notes down below, as well as her Python for AI projects too. Thank you so much for being on the show.

Yeah. Thank you so much for having me here. I agree. And yeah, it was a great pleasure to meet you here on this podcast.

Same.