Machine Learning Engineer Roadmap | Machine Learning Engineer Skills | Machine Learning Expert

you are ready to put four to five hours of study every day then how can you build machine learning engineer skills in six months we are going to discuss that step-by-step road map today machine learning engineer is a very specialized skill and they get paid the most as of 2021 if you're working let's say in google facebook or any top product companies you can earn up to 800 to 900 000 a year as a machine learning engineer and they are more than you know doctors or surgeons so therefore the effort that you need to put is going to be humongous here but in six months you can develop a basic foundation and then after it becomes a lifelong learning journey diagram. I have shown here quickly summarizes the exact steps in your machine learning journey so let's start with computer science fundamentals. I'm assuming that you don't know anything about computer science you are starting from scratch and the first thing you need to learn is computer science basics what is bits and bytes how ram and cpu works the basic fundamentals about internet and so on and for that we have an excellent Khan academy course where you need to follow the first four sections digital information the internet programming and the algorithms by following these four sections you will develop a basic understanding of computer science principles if you are already a computer engineer you know about all these topics then feel free to skip ahead move on to the next one. The next one is level one coding tutorial machine learning engineer has to be a good software engineer first and as a software engineer obviously we need to know programming language. And which is the best programming language for ML engineer no doubt python. Python is the one there are many companies who build their own customized models and for performance reason and for other reasons they want a c plus skills so you have to learn python and learning c plus will be helpful as well for python I have a complete tutorial playlist you can follow that first 14 tutorials and at the end of the every tutorial there are exercises as well a good software engineer has an important skill of a solid understanding of data structure and algorithms in data structure algorithms you need to know about basic data structures such as array list hash map tree graph etc and for algorithms you need to know about you know bubble sort various sorting and searching algorithm binary search selection sort and so on and you need to have understanding of big O notation you know how much memory and and cpu resources a particular program uses in terms of big O complexity because machine learning projects are very compute intensive if you have a good understanding of big O notation it's gonna be super useful. I have created this github page the link of this you will find in video description below so first week first two third and fourth week in computer science fundamental computer coding tutorials python coding especially if you look at this playlist you need to follow first 14 tutorials at this stage so first 14 tutorials will cover concepts such as variables, numbers you know strings and if and four dictionaries functions working with Json etc so first 14 tutorials you need you know about read write files and so on this gives you a basic understanding of programming it helps you write the basic level of coding and a very important thing is exercises so if you click on this particular link you will find many exercises so let's say I want to do exercise on read and write file okay so here let me see so actually the exercises are rather here I'll update the link so if you look at this exercise read and write file and here the exercise gives a description of what the exercise is about and there is a link of solution so you can practice this problem on your own and then verify your answer by comparing your solution with my solution and there are multiple exercises in many tutorials this allows you to practice the code or the concept that you're learning by watching these videos and the same thing is true with data structures and algorithm tutorials as well if you click on this playlist you find this playlist where I have covered you know array, linked list, stack you all the basic data structures and these are all the algorithms now if I look at my collision handling in hash table video for example this video in the description I have an exercise so if you look at this particular exercise here again I give exercise a description you practice first on your own and then you verify your solution with my solution in order to become a good ML engineer you need to become good software engineer first and a software engineer cannot be a software engineer if you don't have a knowledge of at least one database and relational databases are quite popular traditional databases you need to have a good understanding of SQL or structured query language which is used to query or manipulate the relational databases for that Khan academy has a very good course it's a free course where you can practice certain problems on and you can just develop a basic understanding of SQL cuda vancouver Youtube channel also has a nice playlist so you follow all of that in week five and six and develop a good understanding of SQL then you spend week seven and eight in learning some advanced or level two programming concepts such as you know exception generators and so on here we have a link of that kind of academic course so if I open that course you know you'd learn about creating a table inserting a data creating a book list database so when you do this you get to query or you get to practice those concepts so it's a very effective learning you also learn how to aggregate the data you can run those queries Khan academy is an excellent website it's a non-profit, free website where they teach concepts in a simple manner at the same time you can also practice lot of those concepts and then cuda vancouver YouTube playlist is amazing it's a big playlist you don't want to spend time in going through all the videos I think going through the first let me see the first I have mentioned actually first 12 videos here so first 12 videos until you cover joins it should be good enough okay so you're covering group by join select statement unique key primary key foreign key all those basic concepts and then as you go along your way you can learn more advanced concepts then week seven and eight is level two coding tutorials so here I have my python playlist link where you will be spoiling see clearly mentioned 15 to 27 so 15 to 27 is all about exceptional handling class objects inheritance iterators generators decorators multi-threading multi-processing etc these advanced concepts let you write more advanced programs at the end of week 8 you know about little bit advanced programming you know about SQL data structure algorithm you have built a solid base for your software engineering skills now week 9 to 12 you need to spend in learning numpy pandas and data visualization library which could be matplotlib or c bond numpy and pandas are numeric compute libraries these are used heavily in data science for doing data cleaning or number crunching and these are very fast libraries so you need to have good understanding of these libraries and for data visualization you can either use matplotlib or c bond so here are some of the resources that you can use to learn numpy pandas and data visualization libraries for numpy I have a playlist very simple playlist you can cover it up in three to four hours and after that you can move on to pandas pandas my tutorial playlist is long but you can only follow first let's say nine or ten videos and as you practice on more real life projects you can learn advanced concepts such as cross step stack on stack time series analysis etc but in your journey this basic journey of six months just cover first nine videos and that should be good enough matplotlib is used for data exploration as a machine learning engineer you come across a lot of data you want to visualize that data to make appropriate decision to make decision on which model do you want to use for that matplotlib tutorials very simple seven tutorial playlist you can learn about bar chart histogram pie chart all the basic charting and after that you can learn advanced concepts as and when needed you can use kegel exclusively for practicing numpy pandas related problems so here if i go to kegel data sets you will find many data sets okay let's say heart attack and analysis and prediction analysis data set when i go here i get this 11 kilobyte data set for free i can download this csv file now i can do a lot of data exploration and if you want to get some ideas go on to code here people have written variety of notebooks and you can go through those notebooks and study those concepts for example this particular notebook talks about using skxg boot now if you want to just do data exploration and not worry about machine learning then you can search for EDA which stands for exploratory data analysis and you will find notebooks related to exploited data analysis related to this heart attack database for example this person has used pandas for just plotting those tables then some visualization library here I think matplotlib maybe to pop to render a pie chart and variety of charts so going to these notebooks will give you lot of hands on data exploration how as a machine learning engineer you can do data exploration the most important skill you need to have as a ML engineer is statistics and mathematics knowledge now mathematics and statistics itself is a huge fail if you're starting your journey you would be like how much time i should spend on math and statistics i have created a playlist of math and statistics concepts for data science with data science in mind and you can follow that. There is another book called practical statistics for data scientists by Peter Bruce and Andrew Bruce you can read that book now the reason i'm suggesting only two weeks for math and stats is because as i said you can spend maybe two years there are people who do phd in statistics so you can spend five years just studying math and statistics but you don't want to get yourself stuck you know into that into that ocean so you just spend two weeks studying basic concepts you know inferential and descriptive statistics linear algebra, calculus and after that as you're working on your real ML project you can always come back here and refresh your concepts and learn new concepts of math and statistics I find myself learning math and statistics concepts every single day I always do a reading on new concepts right but that should be done in parallel while you are working on your projects stat quest is a very good youtube channel where you can learn statistics after that fifth week 15 to 18 you want to spend in learning the machine learning now when I say machine learning of course I'm talking about statistical machine learning where you will be using scikit-learn in python for doing machine learning you will be building a classification and regression model when I talk about talk about classification decision tree random forest there are so many concepts that you need to go through so basically one month for machine learning should be good enough in this fast pace the six month ML roadmap so math and statistics for data science here is my YouTube playlist the playlist is ongoing as of may 2021 if you are watching this video in future you will find more videos but I talk about basic concepts such as standard deviation normal distribution in z-score and when I have this video I will discuss how exactly uh these concepts are used in the field of data science and machine learning for example normal distribution you can use it for outlier removal using you know three standard deviation and if you look at description there is an exercise link and if I look at exercise I will always give an exercise I mean most of the videos will have exercise along with a solution length so you watch a video then you practice on the exercise which consolidates your understanding of that particular concept and then there is this book I have given the complete book name along with the author name in Amazon here is how that book looks like this is not a paid promotion by the way I find this book to be useful and I hope you can find it useful too and then the here are some free resources on machine learning if you look at this playlist that I have we start with what is linear regression gradient descent dummy variable, hot encoding logistic regression decision tree these are all for classification problem svm random forest k-4 cross-validation k-means clustering for unsupervised learning hypertext parameter tuning using grid so cv l1 and l2 regularization and important thing is the project so after you learn all these video concepts you can build an end to end project for example here I build a property price prediction website for the city of Bangalore where we build a website the whole end-to-end project which we deployed to amazon cloud as well see the last video was Amazon cloud we also built a website so if I look at this website video here see this website you enter all your area whatever and the website when you click on estimate price button it will do price estimation and it will be using a linear regression model that we built in that particular project we have image classification project as well which you can practice couple of feature engineering videos as well and in the videos we'll have exercise for individual topic for example you are watching logistic regression video here we'll go through the concept you know some theory first so I go through theory then there is a code code that we write for that particular concept and in the end we have an exercise so if you open this link for the logistic regression code and if you go to the bottom there is an xss description and you can practice on this exercise to solidify your understanding for logistic progression let's say and I have videos in hindi playlist as well you can spend next one month week 19 to 22 in deep learning so deep learning is all about neural networks you need to learn about computer vision or convolutional neural network models as well as NLP which is using RNN transformer etc and then week 23 to 24 is about ML OPs you need to learn one ML OPs tool such as ML flow. ML flow is one example there are other tools as well so when you learn that you get a full understanding of how you can use to that tool for the whole machine learning life cycle management so here is a playlist I have for deep learning this is this particular playlist is using tensorflow you can use pytorch as well tensorflow and pytorch like you need to use just one of these in here you can see neuron what is neuron here is the simple example of how neural network works then you go through activation functions your gradient decent implementing basic first neural network in python chain rule all these important concepts you know drop out regularization so do whatever you can in the in one month and try to get some understanding of RNN and CNN because those are the two most popular ones again I have an exercise in some of those videos so practice on those exercises and then comes mlops where you need to learn a tool such as MLM flow for example I have given a link of ml floor tutorials so here if I look at this particular tutorial you'll see ml flow is a library you can import in python and then you can use that as a context manager while you are training let's say your elastic nate machine learning model and then you can log different parameters eventually when you run ML flow ui you will be comparing different models with different parameters that performance and you can export those models to pickle file and deploy them to cloud and it just it automates a lot of things that you would be doing otherwise manually so after six months or 24 weeks you would have learned basic or fundamental skills as an ML engineer you need to also know about pi spark hadoop distributed computing because many companies have humongous databases which are distributed in hadoop cluster and you need to use the distributed computing using pi spark docker and containerizations are important concept that you need to have knowledge on CI CD using Jenkins is one other thing that that's an essential skill for any software engineer or ML engineer when you're building a code of course it is maintained on git which is very popular version control tool you also need to know about fast API tensorflow serving etc to build http servers around a trained model and at least one noSQL database understanding and the most popular one there is mongodb now you'll be asking me it's just too much to learn all right just too much but think about this how much how much money ML engineers get okay let's go to a levels dot fi website and here let me search for ML and AI engineer okay and here I'm just sorting google machine learning engineer l8 earns 1 million a year 1 million a year it's just too much money so ML engineers get the highest salary in tech world okay for that reason you need to know a lot of things and the six month roadmap that I gave you gives you just a base you need to build on that base now learning these additional skills or even practicing those other skills that I mentioned before might require you to spend maybe two to three years in order to become a skilled machine learning engineer but the six month roadmap gives you a solid strong base on which you can build additional skills now becoming machine learning engineer is not about technical skills you need to know soft skills and the first one is you need to understand the principles of effective learning you have to learn so many things and there is less time so you want to know how you can spend less amount of time and get maximum output so here is a video on how to learn things effectively you know where you want to spend less time in your input which is watching let's say tutorial videos but more time in output which is reflecting implementing and sharing I discuss various other things such as distraction free learning for example in your 4 to 5 hours of study you will keep your mobile phone away mobile phone cell phone is the biggest distraction another important tip is a group study so if you click on this link and if you go on my discord server here you will find a partner and group finder chat so in the discord server if I go here partner and group finder here you can make buddies you know you can form groups with people and do the study it's like going to a gym if you go to gym alone you will not be motivated but if you go with your friends you will be very much motivated so use this partner and group finder discord server link to form a group and do the studies together so that you can hold each other accountable and then following disciplines and not giving up while there is lot to learn you need to have a faith that you can do it you have to spend your time in a disciplined way do a lot of hard work and it is certainly possible world is full of opportunity the future of ML engineer is very very bright so I hope you can follow this road map and if you have any other question post in a comment below. I wish you all the best

Channel: codebasics

Views: 50,885

Rating: 4.9365706 out of 5

Keywords: yt:cc=on, machine learning engineer roadmap, machine learning engineer roadmap 2021, machine learning engineer, machine learning experts, machine learning expert, machine learning roadmap, machine learning roadmap 2021, machine learning for beginners, roadmap for machine learning, expert in machine learning, how to become a machine learning engineer, machine learning, ml engineer, ml roadmap, roadmap to machine learning, complete roadmap for machine learning

Id: T4MLrtOKPjY

Length: 22min 25sec (1345 seconds)

Published: Tue May 11 2021