So guys, in this specific Article, I will be talking about all the things that you really need to learn in order to master generative AI. ThisThis field has become really popular in recent days based on the amount of research that big tech giants are specifically coming up with. You know the kindskinds of LLM models, LIMmodels, and models, and multi–models that havehave been created. If you see all the tech giants like Google,Meta, and Meta, and X,X, you know which is owned by Elon.Elon. And if I talk about companies like Anthropic, they are in a furious fight to probably develop the best LLM models. Again,Again, they are trying to setset that kind of benchmark. And if I talk about other companies thatthat are really looking toto automateautomate most of the usecases, they cases, they are specifically using all thesethese kindskinds of LLM models or multi-modelsmodels in order to solve the use cases as well. as well. So if you are already learning generative AI, then the plan andand roadmap that I specifically give over here will definitely help you out.
Okay, so if I talk about generative AI over here, guys, two important things you really need to do andunderstand are understand are whatever things I probably say and and whatever skill sets I probably describe.describe. I have already uploaded videos with respect tothis, and this, and every playlist will be given in the description of this particular article. But the way that you should specifically learn it is very important. Now with respect to generative AI, if I probably talk about first of all, I really want to talk about prerequisites.. OkayOkay, so prerequisites areare the most important thing,, and thesethese prerequisites will be very important. If you are skipping this, trust me, it becomes very difficult to crack the interview. If I talk about prerequisites, at least one programming language like Python, okay? You really need to be familiarwith them with them because right now, whatever frameworks and libraries are actually coming up with with respect to generative AI, specifically with respect to LLM models, LIMmodels, and models, and even multi-model, let it be companies like Google and and Meta, everybody is basically working withwith thethe Python programming language. Sothe Python the Python programming language is a must. a must. This is the proper prerequisiteprerequisite that you should really do.The second The second is obviously statistics. Okay, statistics is something that you really need to be good at, okay? Again, the reason why I’m telling you to study statistics and understand how you can apply it in a real–world scenario is just toto makemake sure that you answer all the questions in the interview, right? So one of the favorite questions that are specifically asked in interviews, whetherwhether it be generative AI data science or or statistics,statistics, will definitely be involved. TheThe third thing is with respect to NLP and computer vision. See, now many people like to specifically work with LLM models or LIM models. So that is the reason why companiesare now are now coming up with multi-models, right? So multi-models, okay, if I talk about multi-models, right?
And this will basically be my base. Understand one thing:: if you are really interested inin only workingworking with test, text-definite use cases, right? So at that point inin time, your NLP should be really strong, okay? NLP, NLP, okay? If you are probably planning to work in image generation, videogeneration, and generation, and all thesethese kindskinds of use cases, then your entire computer vision should be really strong, okay? Computer vision should be really strong. Now you decide, like, I have an interest specifically inin NLP. So that is the reason I, most of the time,, actually focus on the LLM kind of modelmodel itself, right? Computer vision——I have the idea;; I know about object detection;; I know about things like that, right? So since we are talking about prerequisites, these are the two different categorizationscategorizations that I really want to do, right? So let’s say if you are interested in NLP, then definitely NLP machine learning is important. In this technique of machine learning, you will understandunderstand what is meantmeant byby word embeddings, whatis meant by is meant by textembeddings, and embeddings, and what the different types of embeddings are. are. I hope you have heard about one-hot encoding, bagsbags of words, TFIDF, then word2vec, averageword2vec, and word2vec, and all thesetechniques, which techniques, which you should definitely know as prerequisites. Again, I’m telling you you really need to know because that will set up your base. Unless and until you don’t know this, if you directly jump to generative AI, you can do it. But again, in the interview, it will definitely hamper. The second thing that you really need to master, as a prerequisite,prerequisite, is NLP deep learning, right? Whenever I say NLP deep learning, whatwhat are things you really need to know over here, right? First thingsthings firstfirst, you really need to know about RNN. And along with that, you really need to know about its variants, okay? All its variants you specifically need to know, like LSTM RNN, GRU, encoderencoder,decoder, and decoder, and everything as such. And obviously, after completing all these things, you really need to know also about Transformers and BERTs, okay??
Understand one thing, guys:: since I have been uploading videosfor the for the past five years,every one every one of the specific topics has been covered in my specific playlist, okay? So in short, when I say NLP, machine learning, and deep learning, you really need to have a thoroughunderstanding of understanding of how you probably work with recurrent neural networks, how you work with all its variants,variants, like LSTM and all, how you, whether you understand attention models or not, attention is all you need, and right now most of the models that you see in the market, specifically with respect to LLM models, or transform on a bird variants itself, right? So all these things you should definitely know as a prerequisite. And again, for all this, I have already created a playlist. Let’s say that you are really interested in computervision. Then, vision. Then, computer vision, CNN, and object detection areare the thingsthings that you really need to know in order to probably crack any interview, right? In order to probably crack any interview, because these are the questions that they specifically ask, ask, If I talk about CNN, the variants of CNN, and RCN and thesethese object detection techniques, right? All those techniques that you specifically have to do areare learning all these things. I personally have a lot of interest with respect to NLP. So that is the reason we have never seen many videos that I have specifically uploaded with respect to computer vision. So again, my area of interest will be this. I will keep on exploring this part. So once you probably complete the prerequisites, okay? And once you understand some of the machine learning algorithms and all, then you come towards learning generative AI, or, you can probably say, yeah, learning gen-AI. Gen-AI. Okay? So, with respect to learning Gen-AI, I definitely like to categorize this into multiple things. One, okay, first, second, okay, I will just try to categorize this so that based on this categorization, you will be able to learn things, okay? What all things this categorization will be focusing on second and third okay now if you really want to learn gen AI or generative AI obviously after covering this particular prerequisites now the popular question is that Krish can I directly jump into this okay you can but in interview will face a lot of problem if you don’t know the basic things okay if I ask you about word embedding you will say hey I will be using Ola my embedding opening and If I ask you, hey, how does it work? How are all these specific models used? So if you don’t know what to make or how these models are used, then you’ll not be able to answer it, right? So that is the reason you really need to be good at this. If I talk about learning generative AI, okay? First things first, you definitely need to know about frameworks. Frameworks to develop, right? Frameworks to develop GenAI applications. Now, whenever I talk about frameworks, there are a lot of things that I specifically want to talk about, so some of the examples with respect to frameworks, like Lang chain, if I talk about Lama index, if I talk about Chainlet, and obviously if you are following me, I’m making sure that I cover all these things. So I will talk about frameworks also; I will write it down below if I talk about frameworks. Right? This framework actually helps you work with any kind of LLM model. So one is the Lang chain. The other one is the Lama index. Right? The other one is Chainlet.
Along with this, you also have a hugging face. Right? So with all these frameworks, you can definitely work with any kind of LLM model. Okay? All these frameworks are super important. So this is one of the things that everybody should basically know about it, okay? Parallel learning, and this is the strategy that I also followed when I probably started a couple of years ago, okay? Along with this, if I talk about frameworks, then the next series is LLM, multi-models, and all, okay? Along with this, when I say multi-model, again, it works with both text and images. Now here, there is huge competition, right? A kind of competition is there among all the big tech giants, right? You really need to understand which performance metrics are better because in a company, you go tomorrow, right? They may ask you, Hey, try to use any kind of LLM model that you have; try to use open source LLM models. So here I will also be writing open source LLM models. Open source LLM models, okay? So for any of these particular LLM models, you specifically need to know how they work. What is the performance accuracy? and you need to keep on researching these things. Not only that, let’s say there is a model as a service, like AWS Bedrock, which provides you with multiple models of access in the form of an API. So you should also know about that and how you can consume it in that specific application. So start exploring all these things. Some of the examples are like OpenAI. As you probably see in my playlist, I’ve created videos on OpenAI. I’ve created video from Gemini Pro, right? Gemini Pro. I’ve created videos from Cloudy. So everything. Why have I created this? Just to make you understand how you can consume all these kinds of APIs, you should probably start working on creating a chatbot application. The third thing from frameworks, then LLM, and then the third thing, the most important thing, is nothing but fine-tuning, right? And if I talk about fine-tuning techniques like Chlora and LoRa, right? And how do you fine-tune with different, different open source models, too? fine-tuning with open source model, and if you know, I’ve also created a video on how to fine-tune with the help of Mistral Lama to write Gamma model everything, so open source model specifically helps you in fine-tuning in an amazing way, right?
So that is what many companies may require in the future: hey, if there is an open source model, please perform the fine-tuning and try to see whether our data is basically able to accommodate it. And the best thing about this open source model is that it can be used for a commercial purpose; only the deployment part is something that you really need to take care of. If you don’t also want to take care of the deployment, then you can use AWS Bedrock services and all. So in short, these are the main three things that you really need to focus on: frameworks; you really need to master frameworks; a couple of frameworks; at least I’ve already created a playlist on Lang Chain; soon I will also be creating a playlist; I’ve created a playlist on Lama Index; and I’ll also be creating a playlist on Chainlet. So there you’ll be able to understand how you can consume different different LLM models and all let it be hugging face Let me open AI Google Gemini Pro right any any any companies any anything as such and that is what is the best thing that these frameworks are specifically doing then I have understood about various LLM models Then I also know about the fine-tuning technique right now all the companies specifically require all these particular skill sets if you want to be a differentiator right now. Let’s talk about a differentiator becoming a differentiator, right? Differentiator. I hope one of the techniques I’ve always stressed in this technique is MLOps, right? MLOps, whenever I talk about having a CICD pipeline, GitHub actions, and all, you also need to think about LLM Ops, right? And this is the upgrade that you really need to do. How these models and APIs can probably be accessed, how this entire fine-tuning process can be automated—you know, how we have to make sure that this fine-tuning process is done in such a way that the model gets updated every time. Inferencing, right? Inferencing techniques. One recent inferencing technique that I probably showed is something with respect to Grok. So we recently heard about this. I have already created a video about Grok. It is an amazing LPU engine. For inferencing purposes, it will provide you with an amazing response. Within less time, you will be able to get the response. Again, it uses open-source LLM models itself. LLM Ops, why am I telling you to focus, and probably, as we go ahead, more of this kind of platform will be coming up, so that the entire life cycle of a Gen.AI project will be handled, okay? The life cycle of a GenAI project will be handled through this, okay? So that is the reason you will see that there are many, many platforms that are probably coming up. One of the platforms that you see is something called Vertex, right? Vertex.ai from Google is providing this kind of platform to do it, okay? So, all these things are probably coming up, but in short, if you really want me to master this entire generative AI, there are two amazing things.
One is a prerequisite. Obviously, I think many of you, if you are continuously following my YouTube channel, at least have completed the prerequisites. For people who are already learning generative AI, I have created frameworks, I’ve shown multiple models, and I’ve shown even all the fine tuning techniques, Laura, and all the theoretical intuition that you specifically require. It’s all about how you can specifically use and as I go ahead You know as we go ahead a lot of LLM ops platforms will also be coming up Which I will make sure that I’ll explore those also and probably create the entire life cycle of a genuine projects at the end Of the day as I said finally Whatever things you run always make sure that implement implement Many end-to-end projects as possible, right many end-to-end projects right end-to-end projects now when I say end-to-end projects one of the most popular project you can see rag Q&A right fine-tuning your own custom data all these types of projects completely needs to be done end-to-end along with the deployment so once you probably do this trust me there is no one to stop you to probably get any kind of jobs I Understand one thing guys.So yes, I hope you like this particular video. This was it for my side. I’ll see you in the next video. Have a great day. Thank you and all the tickets.
