How I'd Learn Python for AI From Scratch (Full Roadmap & Projects)

Indexed: 2026-05-21

[00:00:00]Hey friends, now there is one question that I keep getting. Barra, I want to become an AI engineer, an AI expert, should I learn Python anymore. Now that we have an AI that can write code for me well the short answer for this, yes, Python is the core behind an AI system, every model SDK, every agent, framework, it's all Python. So yes, if you want to get into AI, you have to learn Python.

[00:00:23]But now comes a bigger question is what to learn? Because Python is huge. Now if you go and open any random Python course and you start learning it, you might learn so many topics that are unnecessary for AI. Web apps, machine learning, deep learning, data science, and as well you're going to find hundreds of Python libraries that you never heard of. Stuff that you will never actually use while building an AI system. And this is a big issue if you start learning a lot of things because my friends without a clear focused plan, Python start feeling endless and you start forgetting actually what you have learned at the start. So we're going to go through multiple AI projects that I have done one by one and start extracting the Python topics that you have to learn with that you get feeling how things looks like and why you have to learn them. And of course at the end you're going to have like a list of clear topics that you have to learn for Python. So now let's get started.

[00:01:14]>> Okay, let's get this started.

[00:01:15]>> All right, so let's start with the simplest AI app that you can do in Python. You write a prompt inside the script. You send it to the model. It could be Antropic or OpenAI. And then you get a response from the model back to you as an answer. So this is the Python scripts for the AI app. And here I have a simple prompts. I'm saying okay explain what is a data pipeline. And if I go and run it now look at this. We have a response from the model. It says a data pipeline. It is automated steps that collect, process and move data from one place to another. So actually that's it. As you can see this Python code is very simple. We send a prompt to the AI model and then we get the response back.

[00:01:53]Now, of course, we have a look in order to understand what are the topics that I have to learn in Python in order to write such a code. Okay. So, now what do we have over here? We're going to start with the foundations, but I will not keep repeating that. So, first you have to learn how to create variables in Python. So, the client, the prompt, the response, all of those are variables.

[00:02:09]So, this is one thing. Another one is the data type in Python. So, you have to learn how to deal with the string values inside Python. Prompts are usually string and especially you have to learn how to deal with the f strings in Python because usually we have to construct the prompts using multiple variables. Now since we are talking about data types we have here advanced data type the dictionaries. This is the most important type that you have to master in Python because almost all the response from the API is going to be stored in dictionaries and you have to learn how to deal with it. So dictionaries are always key valued. So we have a role users and the content going to be the prompt. Now here we have another data type. It is inside a list. So the whole thing over here is a list and inside it we have a dictionary. So that means you have to learn those advanced data types in Python. The list and the dictionary.

[00:02:58]Moving on. Another thing that you have to learn is how to import things inside your scripts. Internal external modes and libraries. You have to learn how to do it. So we have here the first three lines. Like for example the first one import OS going to let Python read settings from your computer environments. And another one we want to import the env. This is going to help us in order to read the API keys without storing it inside the code. And the last one and the most important one is how to import let's say a library in order to interact with an AI model. Here I'm using the entropic. Of course you can go and use the open AI or whatever model that you want to work with. So you have to learn how to import things. Another thing that is very important for AI applications is how to deal with the API keys. Now an API key it's like a personal password that you could use in order to interact with the model using your account. So without key without password you cannot go and interact with the AI models. It's very simple. You're going to go to your account and generate a key for your code. Now of course we never put the passwords inside our code.

[00:03:58]You have to learn how to put it inside an environment file. So it's like a hidden and protected file that you can have inside your project. Like for example over here of course I will not show you my access token. It has like a name and then you can have here like a key. So this environment file will never end up like in GitHub by accident. So that's why you have to learn how to get those API keys from the hidden file like I'm doing here. For example, I'm just saying here go and get the key by just using the name. So now another thing that you have to learn is how to send a request to the model. So it's very simple. We are using here create and we are passing for it some parameters. The first one is which model from the introic that we're going to use. the number of maximum tokens that are allowed and then we have the message. So with this we send the request and the result going to be then stored inside the variable response and then we have to go and print it. As you can see you have to learn how to deal with the list in order to print exactly what you need.

[00:04:49]So my friend that's it. This is the simplest AI app that you can build using Python and you have to learn the following Python topics. So the foundations, the variables, the string values, how to work with the data structure, especially the dictionary and the list, how to import external things inside your Python code, how to store and read your API calls, and at the end, you learn how to send a request and receive an answer. All right. Now, another very important AI app that you can build using Python is we build a chatbot for companies. So the main idea is to build something very similar like cloud AI or shad where the customers or the users going to go and start having conversation with the system. So they write a prompt they get an answer and as well very important there is a memory system for the chatbot. And now I know you might say you know what why do we have to build the whole thing using Python we just go and send the customers to shad and they start interacting with it right well no there are many reasons why do we have to build such a app like for example you want to build the chatbot directly integrated inside your system so inside your website or application you would like to have like a window where people directly in your system interacting with the AI and having conversations you would like as well to connect the company's data with the AI model so that you can pull more context to the AI model and ask specific questions about your business and as well we can tailor the system exactly like we need for our business by controlling the behavior of the AI. So my friends there are many reasons why we use Python in order to build AI chatbots for companies. All right. So now here we have a very simple example on how to build a chatbot using Python. So the main idea here is to build a conversation. So you keep interacting with the model until you say stop. So the basic idea is to just have a conversation with the AI model. So we keep interacting with it until we say quit in order to stop. Like for example here I'm going to say what is data engineering in one sentence. So look at this. I got an answer from the model and now I have to ask the follow-up question or let's say keep up the conversation like to say what do you mean with data processing. Now the same thing can happen. We're going to get an answer but look at this the answer is short. So that means it is remembering the history it has here the context about in one sentence. So actually it is working and now I can go and say okay go and quit let's say bye-bye. So this is very simple and yet very important use case that we build in projects. Now of course the question is what I need to learn in Python in order to build such a app.

[00:07:18]Okay. So now what do you have over here?

[00:07:19]Same things you have to learn how to import, how to deal with the keys safely and as well how to send a request to the model and how to receive a response. But now our chatbot it is more advanced than the previous app. Here we don't have only one single request and one answer.

[00:07:34]Here we have to keep things running until something happen. So that means we have a behavior and we have to learn how to control the flow of our application.

[00:07:42]And that's why you can see that we're going to put the whole thing inside the loop in Python. And here we are using the while loop. So we are saying first go and capture the input from the user and then we have to check it using the if statements. So if it has like quit or exit or maybe you can go and add stops then it going to go and break otherwise it's going to go and call the function ask. So we have here a lot of things to learn how to create a loop in Python.

[00:08:06]How to use the conditional statement if else. How to use statements like break continue and so on. And what is more important is how to use a function in Python. So you can see here we have built a function in order to send a request and receive response. And not only that, we are as well building like a memory for the system. So we have like a list and we are appending all the messages inside it. So by looking to this we have for sure to add some fundamental topics in Python like creating a loop, conditional statements and functions. So those are very important tools to build any logic inside applications. They can help you to build a chatbot. Now before we go to the next AI Python app, there is something really important I want you to understand my recommendation as you want to learn Python. The golden rule for me is always 20% consuming tutorials, courses, blog posts and 80% practicing and building. And if you follow it, I'm going to promise you a big difference because honestly this is why most of people feel like they are learning but after the course ended they cannot build anything and they think okay they need another course so they jump to another tutorial and they going to have the same issue. So if you are picking any course or any tutorial make sure it makes you type code not just watching. And this is exactly why I recommend data camp. They focus on building structures that has hands-on exercises going to make things clicks in a way that passive tutorials just never did. Now there are two tracks especially that I like. If you are starting from zero begin with the Python data fundamentals track which going to cover the variables functions list dictionaries which is all hands on with no passive videos. Then you can jump into their associate AI engineer for developers track. And this is the practical bridge between Python into real AI work. You're going to build applications the one that I just showed you like chatbots semantic search engine recommendation system using open AI API hagging phase lang chain also covers LM ops so you can deploy maintain AI systems in productions. Now those two tracks going to help you to work toward the data cam's AI engineering certificate to prove your skills and by the way you can get 25% discount using the link in the description. Data capab thank you so much for sponsoring this video. Now let's move to the next app.

[00:10:19]>> All right next suggest next suggestion.

[00:10:21]>> Now the next very common AI project that we do is talking to documents. So the thing is companies have load of documents, wiki, contracts, policies and they keep updating every day and those stuff was never part of AI training. So the model does not know anything about it. So what we do for that we build an AI system for companies to let the users talk to the company's documents. It's like uploading documents to Shajibet and you ask questions about it. But again for companies we have to build Python application. Now here things get a little bit more advanced because we have to build something called drag. The idea is very simple. You feed your documents into a pipeline that prepare them for AI. So first your documents going to be splitted into small pieces. We call it chunks. Then a small embedding model going to turn those chunks into numbers.

[00:11:10]We call them vectors. And at the end those vectors going to be stored into a vector database. Now when a question comes from a user then the database going to search for the relevant chunk and then hand everything for the AI model. So the chunks and the questions and now the AI model can have all the context in order to response for this question. This is what we call a rag.

[00:11:32]All right. So now let's check the example here. We need more things like for example documents. I have here two documents the handbook and the refund policy. And then we have the whole logic in order to build the application. And inside it I have stored the prompts like for example what is the refund policy?

[00:11:47]Of course you can make it an input in order to get it from the users. Now after I run the application I'm going to get a response like this. So it says refund policies within 30 days of the purchase full refund and available no question asked and so on. But of course using the AI model. So let's see what do we have to learn in order to build such a thing. Now about the imports here we need a new tool in order to build the vector database. Right? And in this example, I'm using the chromabb. Now chromabb it is a small open-source vector database that you could use locally at your PC. So everything going to be stored locally like I have over here small database where we can use it in order to store the small pieces from our documents. So this is only about the import of course after that we have to go and create the database. So here we are using the chromat and as well we are creating something called collections.

[00:12:37]Collections are like tables inside the normal database. So those are the first step and after that we start defining the functions for our process. So what do we have to do first is to chunk the text. So we have to split the document into small pieces. This is one way on how to do it. So this functions get a text from the documents and here we define the size for each small piece as 500. Then we're going to go and build a loop in order to go through the whole document and start slicing the text. We are using the slice mechanism of a list in order to divide the text into multiple pieces. So again, you have to learn and master the data structure in Python. Now once everything is splitted, we're going to return the chunks as a list of items. And now next, we have a very important function that do the whole process. First of all, a very simple loop in order to go through all the documents that I have inside the docs. And after that comes something important to learn is how to work with files using Python. So how to open it, how to read its contents. Now after that we have another loop for each documents.

[00:13:41]So what we're going to do, we're going to run our nice function, the chunk text for each one of them. And after that we're going to loop again for each function in order to split it into pieces using our nice function that chunk text. And at the end comes the most important step. We're going to pass it for the chrom where it's going to take the small pieces and store it inside the vector database as embeddings. So that's the whole process.

[00:14:06]You loop through your folders. You read all the files, split them into small pieces and then hand it to a victor database in order to store it. So those are the stuff that you have to learn in order to build a rag inside Python to answer business questions using the business documents. All right friends, so now let's go and recap and see the whole road map learning Python for AI.

[00:14:28]So the first phase you have to learn the foundations like we saw variables are everywhere. You have to learn the basic data types especially the string because prompts and chunks everything use string data type. Then since we are building applications we have to learn how to control the flow using loops for loop and while loop and how to build conditional statements using lls. And then next you have to master the data structures and all its tools. This is a must especially dictionaries and list and how to deal with the nested structures. So list inside list, list inside dictionary, dictionary inside list and so on. And after that you have to learn how to build functions in Python and how to call them, the different types that we have. So those are all the foundations that you need and without them you cannot do anything.

[00:15:13]And by the way, I have covered all those topics in my 13 hours Python course. It is full of sketches and deep dives. You can pick the one in YouTube or in my data academy. Now after that comes the intermediate level where you have to start learning how to work with the files like how to open a file read it and write into it. And here I want you to focus on one type how to work with the JSON files. JSON is the standard format for many API calls. And since we are talking about APIs you have to learn how to prepare an environment using variables in order to store the API keys and learn the basics about how to call an API and get response from it. And here I would like to add as well important topic how to handle the errors try and accept because we are connecting Python to external resources like the AI models and things can go wrong here like losing connections or exceeding the tokens and so on. And one last important topic here in this phase is the logging.

[00:16:08]So this is the intermediate and then you go to third phase. It's all about AI topics. So how to deal with models from open AI anthropics in order to use the clo code and as well you can learn how to use models from the hugging face if you want free things. So here the main skill is to learn how to send a request and get a response. Another thing you have to learn how to build a rag like I showed you the whole process shunking embedding the vector databases how to store data inside it how to query from it and all those techniques in order to work with the documents. So my friends, those are the core Python topics that you have to learn in order to build AI applications. And of course, there are many other things, but they are not that important like those topics. And as well, the same thing goes for the object-oriented programming. It is not a core topic to learn to do AI applications. All right, my friends. So this is the road map to learn Python for AI that I would follow if I would start over. And by the way, everything is available in the description. So you're going to have a link for the road map and as well for the repository in GitHub where all my AI projects are available for free and I really recommend you to go and do those AI projects in Python.

[00:17:18]They are very simple and you can go and share the success in LinkedIn and tag me there. All right. So I hope you enjoyed this video and thank you so much for watching. I will see you in the next video. Bye-bye.