Running a business from a bedroom - Ask Me Anything (live)

5, 4, 3, 2, 1. All right, we're going live, going live. It says we're live. Are we live? Let me know if we're live because there's always that little interim in between here when it's like trying to connect slash YouTube is trying to catch up with what I can see on my screen. But I'm excited. It is August already, August 2nd. I was gonna do a livestream. I wanted to do one for August 1st, but yeah, some things got in the way yesterday. So we're doing it today. I'm gonna do this stream here to answer questions that I get most often and also questions live on the stream. So if you have anything you wanna ask me about machine learning, about data science, about running a business, about health and technology, about what I'm working on. If you want to just share what you're working on, I'd love to hear it. But it's a beautiful day outside. We're going to spend the next, let's say half an hour or 30, no, half an hour to an hour or so, just chatting, having some fun. Michael, how are you? Abdel Jalali, Jalali. Hello. I'm very sorry if I mess up anyone's name. I'm terrible at reading names, pronouncing names. I'm very, I'm very basic when it comes to words, right? I like to use simple words. It's August 1st where you are. Well, excellent. It's a great month. I'm 26 on September 1st, so I turn 26 in a month. How crazy is that? Let me, let me get some questions ready. We're going to, We're gonna, we're gonna answer some that I've got already. I've got bookmarked. Let me know if you can, you can see this. I might try and dual wield this at the same time so I can see comments. Boom, boom. Is this working? What's everyone up to? Let me know. Otherwise, I'm going to jump into these questions. If you have any more that you want to add, let me know. And we can, we can run away with it. But if the audio is good, if the video is good, if it messes up, leave a comment, I'll get to it as soon as I can. If you can still hear me and see me with this kind of me in the corner view and the screen view, that's all good. Let's do it. I can't believe it's been a month since I left my machine learning engineering job. And it's, it's been, it feels like forever and a day, like working from, from home. But it's good fun. It's good fun. I'm enjoying it. I'm really enjoying it. But without further ado, let's get into it. So we got Hari says, I'm confused between data science and machine learning. Which one should I go for? Please share your thoughts on this. All right, so between data science and machine learning. Well, Hari, that is a great question. So data science and machine learning, I say they kind of, and this is, this is my definition of it, right? It's, they're kind of synopsis, like the same, similar as each other, right? To do machine learning, you need to do data science, and within data science is machine learning. So data science, you're kind of, but, but that being said, right, if you do data science, you might not necessarily do machine learning, right? So I would say in terms of a broader topic, you might have data science and then within that is machine learning. But then some people would also say that you have machine learning and within machine learning is data science. So I would say don't get too confused on— don't spend too much time about worrying which one you should do. I would say do both, right? If you're going to do data science, you'll eventually do machine learning. And to do great machine learning, you're going to have to do data science, right? So that's the way I look at it. I don't try get too— I don't know. I find if I spend too long worrying about like the specifics and definitions of what these two terms are because they're so broad, I'm going to spend a lot of time just arguing what the actual definitions are rather than just working on problems, working on building my skills in either one of them. So beautiful. We can tick off that question from Hari. Now let's go back to the little me in the corner and some more, some more questions over here. We're going to work through these first and then I'll get to the comments. Okay. So Sanjit says, most of the big tech companies like Amazon, Google, when hiring for machine learning engineering positions, test your data structure and algorithm skills. Apart from your domain knowledge. How would one go about learning those if he doesn't have a background in STEM? Excellent question, but again, I'll let you guys know like when I'm not the best person to answer these kind of things, right? Because I only have a limited amount of experience in terms of machine learning. I've only been learning this stuff about 2 and a half years. So I, I don't know everything when it comes to it, and I'll be very honest with you of what I don't know. So the truth is, with big companies like Amazon, Google, Facebook, whatever, whatever big company you want to insert there, I don't really know what they look for in terms of hiring. My goal is not to get a job at a big company like Amazon or Google for now. Who knows how the future may change, but definitely not for now. I want to grow my machine learning skills, grow my data science skills, grow my software engineering skills, and use that to build my own, my own business, my own software products, my own technology products, right? But how would you go about learning data structure and algorithm skills if you don't have a background in STEM? Well, I didn't really have a background in STEM either. I graduated from university with a degree in science, in food science and nutrition, right? So I didn't really do much Science tech, well, I did a lot of science actually, but engineering and mathematics, no, right? How can you do it? The way I've sort of started refining my learning process is to look for like the best book on the topic, right, of whatever topic you want to use, learn. If it's data structure and algorithm skills, get a book and just start to read it because you'll, online, right, you'll have a whole bunch of different courses, you'll have a whole bunch of different resources. And you'll start to, if you start to read a book on a topic, you probably won't understand it to begin with, right? However, the important thing comes from reading it little by little and then researching the knowledge that you want to learn more of, coming back to the book, trying to apply that knowledge and continually repeating that cycle, right? So, as, as for like interview type questions, I think there's a book called Cracking the Coding Interview. I'm sure I've got a copy of it. I never really got into it because I realized I'm— I kind of got a job before I needed to read it. And so, I would look into something like that. Practice your data structure and algorithm skills using that book. Read it. If you don't understand it, that's okay. Reread it again. Keep practicing. That is really the only way that I know best of how to learn something you don't— you you don't really know or you don't have a background in, is to just get a good resource on it. Doesn't matter, almost doesn't matter which resource, but find— for me, I prefer like, like books on topics. I'll read the book almost end to end, and then even if I don't understand on the first way through, I'll just keep going, trying to apply my knowledge, Googling different things, and just keep going from there, right? Now, it's important to remember, learning anything new takes time. So you're not going to learn these things in like a month or, or 2 months or 3 months, right? But, but measure yourself on how much you've learned over like the course of a year or 6 months or something like that. So when you begin, it'll suck, right? You'll suck at it. You won't know a thing about data structures, algorithm skills, etc., etc. But then 6 months' time, you, you reflect back on what you were before you started those gains would be phenomenal, right? So be patient. It takes time to learn anything worth learning. Alright, let's tick off Sanjit. Thank you very much for the questions and thank you guys for tuning in. I really appreciate it. If you're new to the channel, welcome. I make videos on machine learning, health, technology, and basically whatever I'm into, right? Yeah, if you'd like to see anything, if you'd like a question answered, leave it in the comment. I'll get to it as soon as I we can. I'm going to go through these. I'll try speed them up a little so we're not here for forever and I can get to the comments as quickly as I can. Hassan asks, I'm thinking about taking the IBM Data Science Specialization on Coursera. Any thoughts about that? Let's have a look, hey? IBM Data Science Specialization on Coursera. IBM Data Science Coursera. Let's have a quick look at what this looks like because I've never done— there's a lot of courses out there, a lot of good ones that I have never, never done. So IBM Data Science Professional Certificate, beautiful. So it's got data science, machine learning, Python programming, data analysis, data visualization. That's probably the things that you'd want it to have. Where's the syllabus? What is data science? Open source tools for data science. You'll get Jupyter Notebooks, RStudio— I've never used that— Apache Zeppelin— I've never used that— and Jupyter Notebook and Python is probably my tools of choice. Data Science Methodology, Python for Data Science and AI— this is really good actually— Databases and SQL for Data Science, Data Analysis with Python, Data Vis— Okay, this is, this is a good, good specialization, right? You're learning. Then again, a course, right? All courses are pretty good, like on Coursera, all of them pretty good. But the important thing comes from applying what you learn in all this. So you've got, you've got basically all the fundamentals here of data science. If you were to go through this end to end, you would, you would have a great foundation in data science and machine learning. Right, so that's my opinion on that. We go back here. Where is it? I'm thinking about taking— so, okay, so you're thinking about taking it. Any thoughts about that? If you want to get into data science and you haven't experienced before, I would recommend something, a course like this, right? Use this to get your foundations and then work on your own projects. This is really important. Work on your own projects with what you've learned. And what does that look like? You could do Kaggle competitions. You could find a dataset. Work on it yourself. Marry something that you're interested in. For me, I'm really interested in health, so I'd probably look at projects in the health space. If I can improve that with data, improve that machine learning, and then I'd share my work online through a GitHub repository or through a blog post or something like that. That is what I would do. All right, let's go back. Hassan, thank you for the question. Arun, Arun, Arun asked how to be a full-stack programmer. Okay, let's go back. Cam fullscreen. Beautiful. So how do you be a full-stack programmer? I don't actually know, to be honest. Um, I'll tell you how I'm trying to do it. I'm trying to build things that I would like in the world. As in, I'm trying to build applications, or I want to build— I'm learning how to build applications that I want to exist. And what does that look like? So at the moment, I'm learning Swift for iOS development because some of the projects I've I'd like to have, or products I'd like to have, products I'd be a customer of, would be mobile phone applications to do with health. And so, that's what I'm sort of heading towards. And now, how to be a full-stack programmer? I guess if you can, if you can build that application that you want to see in the world, whether it's a web application and you've got things like a backend with a database and whatnot, and then a frontend that people can and interact with, then you're starting to head towards being a full-stack programmer. Now, of course, there's different ways to learn this. You could practice through online courses, but in my experience, right, if you've got the foundations of programming, right, so maybe like a year or 6 months experience of programming, you understand the basic concepts of different functions, of loops, of classes, of structures, et cetera, right? The best way to keep learning on that is to start building your own thing. This is my opinion, right? Is it right? Who knows? But it's just how I learn things best. I need to see things built in the world. So in terms of how do you be a full-stack programmer, you build full-stack applications. And that's how I'm trying to do it. So thank you for the question, Aran. Let's go back. Abhishek. Hey, Daniel. Do you plan to work to gather funding for your startup through the startup school? I saw your previous video and was wondering since you motivated me as well. Abhishek, first of all, thank you for the question. Do I plan to gather further funding for my startup? No, I'm not going for funding. I would prefer to fund it myself. See, I think there's a lot of, a lot of Like, getting funding is good, I think, but being profitable or building something that, that makes money from the start is even better, right? So you'll see a lot of like headlines and news and emails from different startups and whatnot that they've just raised like, let's say, $100 million. But then like, is that, is that success, right? If you're just, if some venture capitalist or some investor is just handing out, handing out millions and millions of dollars to your startup, is, is that the goal you're wanting to achieve, right? Okay, it's great that you can get $100 million, but is that actually gonna, like, is that enough money to build what you want to build, or is that kind of like too much money to build what— how am I trying to say this? If your goal is to build something great, and you require like exorbitant amounts of money to even get started. And I don't know, is that— to me, to me, I think this is me, right? I want to build something that, that attracts revenue from the start so I don't require funding, right? Because I'm kind of, I'm kind of like a— I don't know, I like to do things on my own accord and I don't like having to be tied to someone or owing someone money, right? I prefer— so what I'm— what I'm— how am I going to fund my startup? Well, I'm going to get— spend the rest of the year like working on getting really rich so that I can spend the next 2 to 3 years working on whatever I want to work on. So that's the way I'm planning out doing it, right? Do— is funding a good model? Yeah, but I think, I think there's better ways to do it. I don't have the right answer of the alternative, but I think, I think in terms of creating a business that's profitable from the start is probably better than going hundreds of millions of dollars or getting hundreds of millions of dollars in funding to even before you even make a profit. That's just how I think about things. Get profitable from the start. And again, I can't comment on this because I haven't built a multi-billion dollar startup yet. So we'll see how that turns out over the next couple of years. But great question, Abhishek. Let's go to the next one. Abhishek also asks— two questions from Abhishek, thank you very much. Did you code your website from scratch using HTML, CSS, or did you use a template and build on that? My website is built on Squarespace. And I didn't code it from scratch. I built it before I knew how to code, and I've just left it on Squarespace. Will I build a custom-made site in the future? Potentially, but Squarespace works for me at the moment. Here we go. Eric, how would you combine cybersecurity with machine learning? Um, good question. I do think there was a Kaggle competition for Microsoft Kaggle malware. There we go. Microsoft. So this is how I would combine machine learning with a— or where I would start to look for, right? Microsoft malware classification challenge. And we go here. So this is, this is like where I would sort of start. I don't know how good it is. Like, so this is pretty recent, right? So the malware industry continues to be a well-organized, well-funded market, etc., etc. So this is, this is a pretty good start, I would think, to start figuring out in terms of how you would use machine learning or data science to figure out like what, what malware— does malware exist, right? So this is, this is a form of cybersecurity. This is one great resource you could check out. Otherwise, in terms of cybersecurity, I'd be looking for more examples like this. I don't know too much about cybersecurity. It is very important, just as even more important probably than the machine learning, right? Because if your, if your service isn't secure, then it's probably not a good service. So great problem to have, but this is probably where I'd start, something like this. Let's go back to Notion. We're getting through these. We're getting through these. You having fun? Am I doing the right thing? Let me know if, uh, if you'd like me to do, do anything differently at the moment, but keep, keep the questions coming. I will get to the comments shortly. I'm just going to get through all of these. Wide, WideLytics asks, hi, can you tell us what are the first important steps towards building a successful online business? Great question. Um, so successful online— first important steps towards building a successful online business. Well, the most important thing of any business is to get a paying customer. Without a paying customer, you don't have a business, right? And so the way I think— if I— let me break down an online business, right? If you have, if you're making something, let's say a product or service, your job, right, is to get people, traffic to that product or service and sell it to them. So that's, that's the definition of an online business. Now, how do you do this? Well, you can either make your, and this is, this is what I prefer, you either make your work so great that people can't ignore it, right? There's a book I think called That's So Good You Can't Ignore It, so that they continually stumble upon your work. Like say for example, they're looking for, you're looking for the best YouTube channel on machine learning. It's almost impossible not to stumble upon Siraj, Siraj's channel, right? Because it's incredible. And he's turned that into an online business. And now there's, there's many examples of this, right? You could name, name dozens, hundreds of different online business examples. But the fundamental principle, right, is just making something good that people want to pay you for. Now, what is that? You have to figure that out yourself, right? That's where the value is. That's why people are going to come and pay you, is because you figured out or you've worked out how that you can sell your skill through the form of a product or through the form of a service. And because now it's on the internet and it's, it's the best in class, people are going to come and want to pay you for it. So have something that you sell, drive traffic to the thing that you sell. That, that is, that is online business in a nutshell. So how does mine work, right? So I make videos on machine learning and health and I write articles and so people like to read those articles and watch those videos, at least I hope, And then through that, I don't have anything that I sell directly at the moment yet, but I am partnered with different organizations like Coursera with, with affiliate links. So if people sign up to a Coursera course through my affiliate link, I will get some money from that. Same thing with Amazon affiliates through books. And then if I write a Medium article and if it's published and people read it, and they, they clap for it. Medium have, have a way of a partner program that they pay you some percentage of the membership fees that, that members who read your article and clap, some percentage of that membership fee will go to you. So, that's, that's a few different ways that I, I make income at the moment. I'm working towards having a product to sell. Now, what does that exactly look like? Stay tuned. That'll— that's, that's in the works. It's a bit more of a longer-term thing. But at the moment, my main source of income is from creating things that people want to, to watch and read and see. And then I make money through either claps on Medium, ad revenue on YouTube, or affiliate links. But again, it's the same principle. Have a product or service, drive traffic to that product or service, and then ensure that people want to pay you for that. So that's, that's me in a nutshell. But those are the important steps towards running an online business. All right, let's go to the next one. Ankit, what are your tips in general for students giving any competitive exams for a university selection? Hmm, competitive exams for university selection. Now again, I don't have any experience. I failed my first 2 years of university, and I've— my high school exams were pretty poor because I was too focused on playing Call of Duty on Xbox, right? So, what are my tips in general for students giving any competitive exams for university selection? I would say, hmm, if you want to actually get into university through competitive exams, I would say find someone who who creates the exams and talk to them and figure out what their strategy is. Find the best expert on those exams. It's not going to be me, right? Because I personally hate exams. I know they have a place, but they're just not for me because I'm terrible at them. So, I would say find the expert person who's setting those exams and ask them for their tips, right? It might be a professor, right, setting those exams to make the bar high so you, you He only gets the best of the best students, right? Now, or who else could it be? Who knows? But that's up to you, right? Partner with them to make sure you're either doing the right practice exams or what's the best study strategy for these exams. So that's, that's kind of what I do. Find the best expert on the topic. And this goes for almost anything. If you want to learn something really well, find the best expert you can. And ask some questions. Maybe they'll answer you, maybe they won't, but it's worth a try. Um, so, Ankit, thank you for the question. All right, Max, what type of business do you run online? I kind of answered this just before. Essentially, it's, it's just creating things that people want to pay for or watch. So I create things that drives traffic to it, which on the long term, um, people end up paying for things, and I get some of that money in return. So, that's the definition of, of, of a business, is right? Attract paying customers through having a product or service that people want to pay for. Let's go there. Max is done. Christian, what specific milestones make you feel like your learning is sinking in? How do you internalize and concretize each new concept? Example, being able to simply teach someone what you just learned, only after implementing said new learning in a project or something else that truly solidified the theoretical? Whoa, that is a heavy question. What specific milestones make you feel like what you're learning is sinking in? Hmm, this is a hard one, right? Because as I said, learning new things takes time. So, how do you know what you're doing, right? What are the milestones? I would say a very good skill to develop early on when you're learning anything is to be able to break a large thing down into small things. And now, this is— you almost can't teach it. You have to, you have to figure it out yourself through trial and error of how to, how to break something big into something small, right? Because too often you'll see, like, when you go online and you'll see the best people at what they do. And you'll be trying to learn something and you'll, because you're just starting out, whatever it is, you might be terrible, you probably are terrible at it because you're just starting. Everyone's terrible when they just start something unless you're extreme natural talent, but assume you're not, right? Because I'm not, I'm not, but you might be. How do you, how do you, what specific milestones, right? So you've got to break large things down. So maybe if it's, if it's with learning to code, maybe It's the first time you write your first function, your first custom function for something, or your first custom class. Or maybe it's the first time you work through a project completely on your own. And I mean, there's no such thing, right? Even in my developer job, right, where I'm being paid to be an engineer, I would spend hours researching how to do things and working on a project. So, don't go looking at the people who are the best at something and then comparing them to you. Compare you to where you were a month ago or 3 months ago or 6 months ago. These are the kind of milestones you should be, you should be considering, right? Now, it's hard to plan, especially with learning, right? Because it's never a straight line, right? It's never— learning never goes like this. It's more like, like around and around and around and around, right? So, it's up, down, up, down, up, down. So, it's, it's really hard to plan out, okay, after 6 months I want to be here, right? You don't really know exactly where here is because along the way you're going to be learning different things and here might actually change to be over here. Okay, so, the only really ways like you can know when you're hitting milestones is kind of in retrospective. Alright, so, maybe you're starting something new, day 1, terrible at it. But your goal, you sort of know where you want to be. So maybe from starting from zero, you want to be a data scientist, right? So that's your, that's your ultimate goal, getting a data science job at a company, right? How does that path look? Well, it's almost impossible to plan it out step by step, but you can go, okay, I want to be a data scientist. Today I know nothing, so I'm going to spend the next 2 weeks figuring out what a data scientist does. Okay, beautiful. 2 weeks' time, I figured out sort of what a data scientist does. Now I compare me to 2 weeks ago. I didn't know anything about a data scientist, but now I know sort of what they do. Okay, they know they do this one thing. Data scientists need to be able to clean data. Okay, I'm going to learn how to clean data. Today I know nothing about cleaning data, but I'm going to spend a month trying to figure it out. Month's time, sweet, I've got some experience cleaning data. Look back on yourself a month ago. Wow, I've just hit a milestone. A month ago, I didn't know a thing about cleaning data. Now I know 6 things about cleaning data, right? So that's— have a big overarching goal, right, that you want to eventually reach. Figure it out steps along the way, but the specific milestones can only be found in retrospective, right? So it's like, oh, and because if you try to compare yourself to to that big audacious goal, right? It's so far away or it's so big that it's like, oh, far out. It feels like it's so far away. But remember, you've actually done a whole bunch of steps to even get where you were, right? As long as you're making progress. So compare yourself retrospectively rather than— what's the future term? Interest? No, I can't think of the right word, but for the future, right? So hopefully that answers it, right? Now, again, he asks, Christian also asks, "How do you teach someone something you just learned?" Well, there's no good way to do this either, right? It's just to practice teaching. Publish your work so you, so public, so the, so that people you're trying to teach can give you feedback on that, right? There's no better way to get feedback on seeing how good your teaching skills are than to try and teach someone. But otherwise, yeah, I think, I hope that that answers your question. At least that's what I try to do, and I'm not perfect at this. I'll look at someone who's really great at engineering in machine learning, I'm like, far out. I used to get really jealous, right? And I still kind of do, but now I get inspired. I see someone really good at what they do and I go, wow, if they can do that, I bet I can do it. So that's, that's my mentality around learning things. Okay, let's go back. We got one more question. Best practices of deep learning deployment. Great question from Ganesh. Thank you very much. Now, this is also deep learning deployment is actually also still very, like a very hard thing to do. But there is, there is one resource I've been looking at. Let me, let me get this, let me get this screen on. Here we go. Full stack deep learning. So I was looking at Oh no, that's not it. This website here, I was looking at this recently actually. And now deploying a deep learning model is probably one of the skills gaps at the moment, right? So you— so courses really focus on building different models and, and exploring data and whatnot. But deploying it is probably a little— a gap in the market, right? Is how do you— once you've built a model and you've collected clean some data, how do you deploy that to thousands and millions of users? Still a hard thing to do, right? So go to this website, Full Stack Deep Learning. This is what they cover. So you've got these books here, which usually cover deep learning, math theory, computer science. Training deep learning models is, is all these amazing courses like Keras, Fast.ai, deep learning. Shipping deep learning projects is what this course covers. I personally haven't done it, but I've been looking at it and I want to do it, and because right now I, to be honest, I don't know the best practice of deploying a deep learning model. I know it has something to do with Kubeflow and Kubernetes. I think that's probably the route you want to go. Package everything you have in a Docker container, put that on a Kubernetes cluster. Retraining is probably a bit of a problem. If you want to retrain your model, you're probably looking at something like SageMaker. I don't know what the Amazon SageMaker that is. I don't know what the Google equivalent is. I'm sure they have one. I'm sure also Microsoft have one as well. But this is probably a course you want to look into if you want to deploy a deep learning model. Full-stack deep learning. What could you— what more could you ask for? So we got formulating the problem, estimating project cost, finding, cleaning, labeling, augmenting data, picking the right framework and compute infrastructure. That's something you'll probably want to know for deep, deploying deep learning models, troubleshooting training, ensuring reproducibility, deploying the model at scale. And again, remember, these technologies are constantly changing, right? So this, this course is the start of 2019. It was done in March, so March 2019, so it's pretty recent, but likely it'll change by March 2020. So Just remember, things you learn today might not be helpful in the future, but that's all right. It's, it's, it's all a learning curve. Every day is day one. So I hope that answers that question around best ways to deploy a deep learning model, which is kind of, I don't know. Let's, uh, let's jump into the comments. Thank you all so much for, for joining in, by the way. I really appreciate it. So we got You got some hellos. Esteban, hey, how are you, mate? Good to see you. Hi, bro, I'm taking the deep learning course on Udacity for your advices. I'm not sure if that was a question or just saying hello, but all the best with the deep learning course on Udacity there. Michael asks, how do you gather requirements from clients? Hmm, I'm not sure what you mean about requirements from clients. If you're talking about how do I How do I— how does a project with a client start? Usually it involves some liaison with someone from another company, and we usually talk it out to figure out the exact problem they're trying to solve and where my skill sets can fit in. So we have Andre, how easy is it to make ML models in Swift for iOS applications? Core ML is actually incredibly easy, and same with Firebase ML Kit. I'm getting really, really into those. I'm learning, I'm learning more and more with Swift, Core ML, and Firebase ML Kit every day. And I think it's going to get even better in the future is to, is to use TensorFlow, like Swift for TensorFlow, right? So I think this is This is something that's just been released in the TensorFlow ecosystem. Swift for TensorFlow. This is what I'm really excited about actually as well, because what this means is that you can do end-to-end product development, uh, as in build the actual product you want to work on in the same language as, as what you're doing the deep learning and machine learning on. So Swift for TensorFlow, this is why I'm learning Swift, is because it's going coming out for TensorFlow. Next generation platform for deep learning. All lots of different exciting things. And because it's still early days, right? When I started learning machine learning, Python was, it's still taking off, right? So it's not a bad thing to be learning Python. Python's actually the language of deep learning and machine learning now. But because it's early days with Swift, it's kind of exciting because if you build something with it, you'll be potentially, or you will be one of the first. So there's a lot of opportunities to contribute to open source, build things in Swift that no one else has built in Swift. So that's, that's a fun thing. All right, let's go back to the, to the comments. Okay, Esteban says, what is the next step when you have the knowledge about deep learning? And it really— so that question is kind of vague. If you want to Guys, I can answer questions, more specific the better. I mean, next steps for doing certain things is, is kind of really broad. So, it's like that could be anything, right? And in terms of next steps about knowledge of deep learning, do you have knowledge of the math or do you have knowledge of the code or what's the knowledge there? So, specific questions are the best ones I can help out with. How do you get clients? Lethal. At the moment, business comes to me. So, because I make things online, a lot of things online, I'll get emails and LinkedIn messages and whatnot of people wanting to collaborate every so often. And I kind of have to pick and choose who I partner with, right? Because of course there's only limited time in the day. So, could I do this better, right? I could do this better in sort of going towards who I want to partner with, so reaching out to them. But at the moment, business comes through to me. And I think, in my opinion, like, oh, actually, I think that's just a better way, full stop, is to just to have business come to you. How do you do that? Again, it's the same principle. Make something online that people want. That, that's the number one, number one thing for online business is just make something that people want. And I think that's why Seas Y Combinator's motto, make something people want. So that's how I get clients. I make things online, they find them, they reach out to me. For example, when I partnered with Coursera, I was making videos about learning on Coursera. Coursera decided to partner with me, sent me an email, we worked out the logistics, and we got a partnership on board. So there we go. One second. Mm. How good is coffee? How good is coffee? Okay, here we go. We have Hashit asks, how do you like this free time you are getting nowadays? Hashit, it is amazing. I really enjoy the fact that I don't have to commute anywhere. I've kind of started living my life in a 3-kilometer radius, as in, I live on the internet now, but the places I go physically, uh, I have a café down the road. My jiu-jitsu gym is down the road as well. There's a beautiful section of water down about 5 minutes walk away. That's where I go for morning walks. I take my puppies to play. Um, but in terms of free time, it still feels like there's not enough hours in the day, even though I've kind of reduced the amount of time I have to be somewhere by like 3x, I still can't find enough time to do things. So, what do I need to do? I need to get better at saying no to things. I'm slowly getting better, but I still need to get better at saying no to things and just work on the things I need to work on. Imran, yes, I did. I shaved this morning. I look fresh and clean. Um, so we have Noah. Do you have co-founders or a team or working on your own at the moment? So my, my online business that I'm running is me, my, just me. I edit all my videos, I write all my articles, I learn all the things myself. Um, my health tech company I'm co-founding with, with my brothers. At the moment. Michael, the audio appears to be a little out of sync. Oh great. Hopefully this fixes itself because it's always hard to fix these livestream things as they're going because it's, it's live. Imran says it's laggy. Hopefully that lag is fixed up as well. Noah, where do you generate the cash flow to maintain your life expenses and fund your startup? I kind of answered that before, Noah, but the main income sources that I have at the moment are affiliate links, right? And medium income and ad revenue from YouTube, which is not much. But the important thing to remember is, okay, you can, you have no, you have no limit to your earning potential, right? But a real, a real like quick and easy way is to assess your expenses. So right now I live in, I live at home with my parents. I've minimized my expenses to be very small. So the amount of money I do earn, kind of, I can use a lot more because my living expenses are so low. So my basic living expenses are food, rent, like AKA bills and that sort of stuff, and online like resources, right? So that's basically my, my living expenses. And so I've minimized that to, to be as small as possible. It's about $1,500 Australian dollars per month. And so I earn more than that per month, right? And so that's, that's, that's an important thing, right, is to, to spend less than what you, what you earn, or reduce your living expenses as much as possible. And now also says, when and how did you get started in programming? I want to build high-level scale web apps. Any helpful tips for that. I have never built a highly scalable web app, but when did I get started in programming? I got started 2 years ago on Treehouse. I started learning Python because I wanted to learn machine learning and deep learning. And how? Through Treehouse and all, all different online resources. That's how I've learned, all completely online. Noah Scully asked, what is the step-by-step process to building a highly scalable web and mobile app? What technologies would you use? I would use nothing but cloud. For a mobile app, probably something like Firebase. I would have the database on Cloud Firestore. Um, and yeah, I would— I think you can use Firebase for basically everything on a mobile app, and it will scale infinitely to number of users you have. It will, it will put in, uh, different servers around the world, so no matter where your people are in the world, they'll have low latency. So Firebase and cloud technologies. As for a web application, you definitely want something like, you got no other option than to, well, you do have other options, but unless you're getting like 10% of internet traffic, use Azure, AWS, or GCP. Use those services. You need a database, you need scalable compute, and preferably these things will scale automatically so you don't have to involve. Be involved in scaling that. Vinod, hello sir, I am learning machine learning from the internet one year. I now have— I know internet level of machine learning. Should I join Coursera course now or for advanced learning? Um, if you've been learning it for a year, I'd say start working on your own projects. Uh, that's, that's where you're going to get the best learning. Don't mistake more course certificates for skills. Right? You want skills over certificates. Now, skills are kind of like a way of showing, hey, I've got these skills, but I think the technology field prefers to see things that you've built. So, build something if you've been doing it for about a year, and I'm a bad example of this. I need to build more things. I'm trying. Now, how did you get your old job? I— here's how it went. I was making things online and I was publishing stuff on LinkedIn. LinkedIn is actually probably the best platform you want to be on for job searching. That's just my experience. And someone reached out to me and said, "Hey, Daniel, I'm really liking the things that you're publishing online." Alright, and so I'm like, "Sweet." And then they're like, "You should meet Mike." And I'm like, "Okay." I met Mike, and Mike was like, "You should meet Cam." And I'm like, "Okay." And then I met Cam, and Cam worked at Max Kelsen, which is where I worked. And Cam said, "Hey, you should come in." I told him what I was into. I want to learn AI. I want to learn machine learning. I want to use it for health. I don't have a degree in machine learning. I don't have a degree in AI, but I do have a degree in health. I want to work out how to put them together. That is my long-term goal. It still is my long-term goal. How do I do that? I don't actually know yet. I'm trying to figure it out. But Cam said, do you want to come in on Thursday? And I said yes, and I went in on Thursday. I did some exploratory data analysis, exploratory data analysis with Pandas, and then I did that again the following Thursday. So I was going in one day a week, and then the third Thursday I got offered a job, right? And this is because I was kind of showing how interested I was in, in learning, right? So of course I wasn't the best data scientist, machine learning engineer, but I was ferocious with what I was learning. I was the one who's like, yes, I can do this problem, whatever it is, I'll figure it out. So that was my attitude. I just believed I could, I could get in there, and it, it kind of worked out. I don't know, I didn't do a formal interview. I didn't do any code on the whiteboard. I never handed in a resume. It was purely from what I was publishing on LinkedIn. And I wrote an article on that. If you Google Daniel Bourke start AI ML, that's— you'll probably find it. Or Daniel Bourke, I want to start artificial intelligence, where do I start? That's the article with the full story. And Gabrielle, what is your business? Gabrielle, I make videos and I make— and I write articles at the moment. And so that drives traffic to different products, and I get money from that. And that's, that's how I'm running it at the moment. And I also have— I'm partnered with DataCamp and A Cloud Guru and Coursera. So I'm making educational content for those platforms, and I'm getting paid through that. So there's a couple of different revenue streams at the moment. Paul, damn, I thought it was 18 hours, man. I'm sorry. I think every time I schedule one of these, I get something wrong. My bad, my friend. It'll be here. It'll be here. Uh, it'll be here for life. It'd be great to look back on these live streams in like 40 years to see if I've actually held true to my word of building something fun, building something valuable. Okay, what do we got here? Carl, what is my new business? I've answered that question a few times, Carl, but thank you for the question. I make things online. I'm stepping away from consultancy. Consultancy doesn't scale. I want to build things that scale, right? And that's code and media. Consultancy is a lot of fine-tuning work for very specific problems, whereas I would prefer— I prefer— see, I'm pretty lazy. I want to make things on the internet that scale infinitely. Simasuka asks, how much math, how deep do you need to know in order to learn ML? This is a very hard question to answer. If you want to do machine learning, you don't need math. You can do it with pure code. If you want to get really deep, you need a lot of math, right? So if you want to do applied machine learning, you can do that with pure code. If you want to get into the nitty-gritty stuff right? So if you want to maybe build a new machine learning algorithm, knowing the math is basically required. I have done no math courses apart from Khan Academy since high school, and I finished high school nearly 10 years ago. So that's how I did math for machine learning, looked up Khan Academy, linear algebra, matrix manipulation, calculus, etc. I just went through those. On Khan Academy, YouTube. 3Blue1Brown, great YouTube channel on math. Paul Garcia, are you going to be partnering up with others or you're building the business 100% yourself? Paul, right now it's, it's purely myself. That's how I get my cash flow. At the moment, I'm trying to— that's what I'm planning to do for the rest of the year is making a lot of money and using that to fund my longer-term health technology business. So that's basically where I'm at is I need to generate a large amount of cash flow so that I can spend the next 2 to 3 years working on whatever I want. Muhammad, still waiting for jiu-jitsu vlog. Yes, great idea. I will get a jiu-jitsu vlog happening. Oh, exciting thing happened actually. Anyway, I can't find it. Don't worry. Um, I got graded again, so I'm now 2 tip. 2 stripe white belt. Exciting, right? And it's not about the grades, it's about getting better at jiu-jitsu. So that's what I'm focused on. I really, I really love it. I'm gonna go train tonight. Kevin, you are Australian, right? Yes, I am Australian. What do you think about the demand for data scientists in Australia? So if you're not relocated in Los, I thought of, what do I think about the demand for data science in Australia? The real answer is I don't know, but I think Australia is moving kind of slow. I did go to an AI meetup last night and the talk was basically about Australia being slower than the rest of the world in terms of machine learning and artificial intelligence progress. Now, again, I'm pretty oblivious to these things. I kind of just like to work on the things that I like to work on and not really pay attention to trends in Australia or other where. All I know is that Their companies are slowly moving towards being able to use data in their business. So there's a big hole there, I think, for someone to come into large companies and basically— this is what, what we did at Max Kelsen, right, where I was working. We went into large enterprise and government companies and essentially showed them where they could use data to improve their business. But there's still a large gap, like large businesses move pretty slowly whereas we could move pretty fast because we were just a team of about 12 machine learning consultants. Um, but really there was— yeah, really, yeah, so there's about a dozen of us. And versus like companies with thousands of employees, right? So it's like you go to talk to one person and then they would have to talk to someone else and then they would talk to someone else. And before you know it, it's 6 weeks after you actually spoke to them before you get like a next step. So these things happen slowly. But I do think there's, there's a big demand in, in, in many different companies of how you're going to use data. Like, it's just part of the software stack now, right? So if you're building, if you're building technology, data science and machine learning is just part of the stack. And here we go. Paul Garcia, what do you think about reinforcement learning? I've never actually seen it do anything outside of like, um, play, play games. Uh, is that valuable? I think so, like in terms of step by step building it up to be something worthwhile. But in terms of deploying it in like an actual production system to, uh, like application of some sort, I haven't actually seen it. Um, there's a fun project actually. My, my, one of my colleagues, ex-colleagues I was talking to last night said that they're kind of trying to work on something of getting reinforcement learning into production. And I can't really— I don't think I can share the exact details, but they're trying to actually do that. I can't off the top of my head think of any production examples, like, I mean, application going to millions or thousands of users that actively uses reinforcement learning. Maybe I'm wrong. Paul, Where you were learning machine learning, how much time was spent learning and how much time you spent working? Um, that's a hard one. So we had a rule called like 20% time. And when I was working as a machine learning engineer, and I still kind of adhere to that, right? So it was like 80% of our time was spent on projects. And then 20% of our time was spent learning whatever we wanted about machine learning, right? So you never know what could have— what you could have learned could have been applied to a project. But even then, within that 80% time of working on a project, it was still a lot of learning, of like researching of different ways to do it, like read papers, like going through online blog posts, looking through the documentation. So it was still, even though it was like 80% on client projects, a lot of the time you're trying to figure out what's the best way to do it for that client project. So, a lot of it is learning on the job, right? So, I hope that answers it. So, it's kind of— there were— put it this way, there was rarely a time where you would go from end to end in a project and not learn something completely new along the way. Terapat asked, what kind of project that make you get a job? That's a very good question. I would say the best way to do it is Find the ideal job that you would like. Find a way to figure out what project that role, what projects that role requires, and then do those projects on your own if you can before even getting the role. Alright, so if someone in an interview asks, "What have you done?" You can say, "Hey, I've done this actual project for this role. Here's the example." And that's going to involve some research from yourself, but that's an important step into getting where you want to go, is researching what it is. Sai asked, what's on my board? What's on my board today? All right, let me read it out for you. My board, we have writing 1 hour, reading 1 hour. That's a given every day. I do those 2. Workout, that's another given. YouTube livestream, we're here. Then we have— we're doing Swift. We're gonna learn Swift for like 4 hours. I have a Y Combinator meeting this afternoon, and then I've got to respond to some emails. I'm getting slack with my emails, and then yeah, that's about it. Workout, write, study, cook dinner, stuff like that. Simple, right? I keep it similar every day. Jenakai asks, what, what should I do after Coursera ML course? If that's the Andrew Ng one, I would start to learn Python, right? Get into coding as soon as you can. That Andrew Ng course is, if it hasn't changed, it's taught in Octave and it's very theoretical. Start getting, uh, getting, getting some Python experience in machine learning as soon as you can. Rishi, how should a person with 11 years experience in IT go about switching career in ML? Rishi, I would start to learn ML and then start to create things with your experience in IT around machine learning. So I would— that's, that's my, that's my best advice for that. Muhammad, congrats, keep it up. Thank you, Muhammad. I really appreciate it. Jose, brother, I'm a fan of your channel. Inspiration for me to take up my GCP. We'll take all. By the way, I'm BJJ Panamericano and I'm training for the Brazilian World Cup. Oh my, my boy! Oss! That's really cool. I appreciate it. And I appreciate the support, man. I love BJJ. I love jiu-jitsu. It's taught me a lot, actually. It's taught me a very valuable skill of being able to lose, right, and then move on as soon as possible. Because that's what happens in jiu-jitsu, right? You'll wrestle someone and they'll beat you, right? So, in the wild, if you lose, if you lost that fight in the wild, you would die. So, jiu-jitsu is kind of very humbling in the fact that you can go up against anyone and then get immediately, well, get tapped out, right? So, you can get choked out, you get joint lock. In the wild, if you get tapped out, you're, you die. So, it's very humbling to lose on a consistent basis, but then just get straight back into it. And you might tap someone out, right? So, you— and then you realize, okay, this is what it feels like to win. So, it's a very important skill, I think, or skill— what can you say? Virtue? I don't know. Or being able to lose, but then move on, right? So, admit where your faults are, figure out what your weaknesses are, and then move on. So, jiu-jitsu teaches you that very quickly. So Kevin says, thank you for these videos. Thank you. Thank you, Kevin. I really appreciate it. Ryan Daniel, in the link I posted above, it is Australian alternative to Y Combinator. Ryan, I don't think, I don't think links come out in the, in the channel, in the chat. I don't think you can post. I think YouTube blocks links from being posted. So if you want to if you want to send that to me, you can find my email daniel@mrdbourke.com. Otherwise, leave a comment on the video once this is live. But at the moment, I don't think YouTube allows you to share links in the live chat. But I think we've been through all the questions unless anyone has one more. We could probably, we could probably stay around for one more, but otherwise I've got to go. I've got to go do some learning. I've got to go do some reading. Some writing. I've got to get an article out for the week, and then I've got a meeting at 1 PM with my Y Combinator team. I've got to make some progress on that. It's an exciting week, right? It's just, there's an endless thing of things to do. This is my list for today. I've tried to make it small. There we go. I'm trying to make this smaller, right? Work on less things, but better things. We got Sai. We'll do one more question. So one more question. When you were studying initially, have you ever had interruption for a week or more? What did you do to get back again? Yeah, of course. Life gets in the way, right? So, um, these things happen. I think it's important to remember, like, you'll see people and even myself who like say, yeah, just keep working hard, keep doing things, keep— no excuses, yada yada ya. But reality is things happen. Life happens. Things get in the way. You know, my car broke down the other day, I need to get it serviced, etc., etc. So these things happen. Important thing is to remember, like, okay, you take a break or something, you can just resume back to what you were doing whenever you really want, right? That's, that's just how these things go. So, if you need to take a break, if you need to take care of something, something comes up, you're sick, you need to improve your health, something happens in your family, these things happen. It's life. What you do, you take care of what you need to take care of, then you get back to what you were working on. So, I think there's no real, real better formula than that. You just take care of what you need to take care of and then get back to what you're working on. There's no, no set way of how to do it. That's just how to do it. All right, so let's, uh, let's rock and roll. Well, okay, we'll do one more. Michael asked, am I crazy spending my runway to invest more time positioning myself for an AI or machine learning role, or should I suck it up and find my next role gig before quitting? So Michael, I I think it's never a bad investment to invest in your education, right? So, as long as you're, as long as you're, you're using that your runway funding to invest in yourself for future skills, it'll kind of look like this, right? You'll have, this is, you got a graph and you'll have you, your results are going here and but your learning is going like this. So, results don't really follow learning exactly. So, learning increasing, results going flat. But then you get to a point where the results will start to catch up to the learning because you'll start to figure out how to apply your knowledge. Alright, so learning might happen quickly, right? But results are still lagging behind. So, remind yourself of this and just keep learning and eventually those results will catch up to it. So, is it stupid? I would say don't be, don't be silly about it. Don't like waste heaps and heaps of time and time and money. Take care of what you need to take care of, learn what you need to take, what you need to learn, and don't forget, you can have all the drive in the world to learn something, but you still need gas, and that comes in the form of having somewhere to sleep and having somewhere to eat. Alright? So, save some money for that. Otherwise, Thank you all for joining in. Welcome to the channel if you're new to the channel. I've got to go do some, some actual learning and stop, stop talking so much. So hopefully, hopefully this video was, was helpful. If you have any other questions, feel free to contact me whenever you want. You can find all of my contactable stuff on my website, mrdbourke.com. Otherwise, there's probably a bunch of links in the description or email me directly, daniel@mrdbourke.com. Otherwise, have an amazing day. Enjoy August. I'll see you in about a month. I'll probably be, I might be 26 by then, which is scary. No, it's not. I'm excited. It's a beautiful day. Have a great day. Oh, as always, keep learning, keep creating. You got this.