What Skillsets Takes You To Become a Pro Generative AI Engineer #genai

Summary

In this video, Kish Naak discusses essential skills for becoming a generative AI engineer, focusing on understanding large language models, image models, and multimodal models. He emphasizes the importance of learning both open-source and paid models, exploring frameworks like Lang Chain and Llama Index, and practicing with various projects to gain expertise in the field.

Takeaways

• 😀 The video is aimed at individuals interested in generative AI engineering and provides insights into the skills required for the field.
• 🔍 The speaker is Kish naak, a data science educator, who discusses the importance of understanding generative AI for those transitioning into the field.
• 📚 The video offers a comprehensive list of resources and playlists for learning about generative AI, including end-to-end projects and tools.
• 💡 Generative AI involves creating new content based on context, with a focus on large language models (LLMs), large image models, and multimodal models.
• 🏭 The video highlights the competition among tech giants like Google, Microsoft, and Meta to develop the best LLMs and image models.
• 🛠️ It emphasizes the importance of learning both open-source and paid models for generative AI, as well as understanding their advantages and disadvantages.
• 🔑 The speaker mentions frameworks like Lang chain and Llama Index as essential tools for developing applications using generative AI models.
• 📈 The video stresses the significance of practicing with various use cases and understanding the deployment and scalability aspects of generative AI models.
• 📝 There is a strong focus on the importance of fine-tuning models with custom data, which is considered a crucial skill in the field.
• 📚 The prerequisites for entering the field of generative AI include knowledge of Python, machine learning, deep learning, NLP, and advanced concepts like RNNs and Transformers.
• 🛑 The video concludes with a roadmap for becoming a generative AI engineer, urging viewers to learn the basics and practice with projects to improve their skills.

Q & A

What is the main focus of the video by Kish naak?

The main focus of the video is to discuss the important skill sets required to become a generative AI engineer and to provide necessary materials and resources for learning about generative AI.

Why is Kish naak making this video?

Kish naak is making this video because many of his students who have transitioned into the data science field are getting work in generative AI and are using large language models and large image models to solve various use cases.

What are the two main types of models discussed in the video?

The two main types of models discussed are large language models (LLMs) and large image models, with a third type being multimodal models that combine text and images.

What is the main aim of generative AI models?

The main aim of generative AI models is to generate new content based on any given context.

What are some of the companies mentioned in the video that are in competition to create the best LLM models?

Some of the companies mentioned are OpenAI, Google, Microsoft, and Meta, all of which are competing to create the best large language models or large image models.

What are the two important categories of generative AI models discussed in the video?

The two important categories are open source models and paid models, which the video suggests one should have a complete understanding of both.

What is AWS Bedrock and how does it relate to generative AI?

AWS Bedrock is a service that provides APIs for various generative AI models, both open source and paid, allowing users to solve business use cases and perform fine-tuning without worrying about the cloud part.

What are some frameworks that one should be good at for developing applications in generative AI?

Some frameworks mentioned are Lang chain, Llama Index, and Chainlink, which provide tools for various functionalities from data injection to transformation and the ability to call both paid and open source models.

Why is understanding vector databases important for generative AI?

Understanding vector databases is important because they are essential for converting text into vectors, which is a key process in developing applications related to text in the generative AI field.

What is the importance of fine-tuning custom data with LLMs in the context of generative AI?

Fine-tuning custom data with LLMs is crucial as it allows models to be adapted to specific use cases and business requirements, making it a vital skill for generative AI engineers.

What is the prerequisite knowledge required to start learning about generative AI according to the video?

The prerequisite knowledge includes Python programming language, basics of machine learning and NLP, deep learning concepts, advanced NLP concepts like RNN, LSTM, and Transformers.

Outlines

00:00

🌟 Introduction to Generative AI Engineering

In this introductory paragraph, Kish Naak welcomes viewers to his YouTube channel and outlines the purpose of the video. He discusses the growing interest in generative AI among his students who are transitioning into data science. The focus is on the skills required to become a generative AI engineer, particularly in the context of using large language models (LLMs) and image models. Kish promises to provide necessary materials and playlists covering end-to-end projects that utilize both open-source and paid LLM models. He emphasizes the importance of understanding generative AI, the prerequisites for entering the field, and the key skill sets needed.

05:01

📚 Understanding Generative AI and Model Types

This paragraph delves into the specifics of generative AI, explaining the three main types of models: large language models (LLMs), large image models, and multimodal models. Kish highlights the importance of understanding the differences between these models and their applications. LLMs are used for text-related tasks, trained on vast datasets, and are in high demand due to competition among tech giants like Google, Microsoft, and Meta. Multimodal models, on the other hand, combine text and image capabilities, offering a broader range of applications. Kish also discusses the importance of skill sets in generative AI, focusing on both open-source and paid models.

10:03

💡 Exploring Open Source and Paid Models

Kish discusses the necessity of understanding both open source and paid models in generative AI. He compares the current state of generative AI to the machine learning boom in 2018, emphasizing the importance of hands-on practice and exploration. He mentions various companies and their models, such as Meta's Lama 2, Open AI, and AI 21 Lab, and discusses the differences between open source and paid models in terms of deployment and scalability. Kish also introduces AWS Bedrock, a service that combines various LLM models into a single API, making it easier to fine-tune and deploy models without worrying about cloud infrastructure.

15:05

🔍 Frameworks and Tools for Generative AI

In this paragraph, Kish introduces the frameworks and tools essential for generative AI development. He mentions the importance of being proficient in frameworks like Lang chain and Llama Index, which facilitate the integration of both open source and paid LLM models. Kish also discusses the use of Hugging Face for accessing various models and the role of cloud platforms like AWS and Azure in deploying AI services. He stresses the importance of understanding vector databases like ChromaDB and Cassandra for working with text data. Additionally, Kish advises on the practical aspects of creating projects using LLM models and the significance of fine-tuning models with custom data.

🚀 Prerequisites and Roadmap for Generative AI

Kish concludes the video by outlining the prerequisites for entering the field of generative AI. He provides a roadmap that includes learning Python, basics of machine learning and natural language processing (NLP), deep learning concepts, advanced NLP concepts, and transformers. He encourages viewers to go through the provided materials and playlists to build a strong foundation in generative AI. Kish emphasizes the importance of practice and project development to become proficient in the field.

Keywords

💡Generative AI

Generative AI refers to artificial intelligence systems that are capable of creating new content, such as text, images, or music, based on existing data. In the video, the speaker discusses the importance of understanding generative AI for engineers, as it involves working with models that generate new content. Examples include large language models (LLMs) and large image models.

💡Data Science

Data Science is an interdisciplinary field that uses scientific methods, processes, algorithms, and systems to extract knowledge and insights from structured and unstructured data. The video mentions that many of the speaker's students have transitioned into data science, indicating that understanding generative AI is a valuable skill in this field.

💡Large Language Models (LLMs)

Large Language Models are AI systems trained on vast amounts of text data, capable of understanding and generating human-like text. The script discusses LLMs as a key component of generative AI, highlighting their use in solving various business use cases related to text.

💡Large Image Models

Large Image Models are AI systems trained on large datasets of images, capable of generating new images or understanding visual content. The video script mentions these models in the context of generative AI, indicating their role in image-related tasks.

💡Multimodel

Multimodel refers to AI models that can handle multiple types of data, such as text and images. In the video, the speaker explains that multimodel AI is a combination of text and image models, capable of solving use cases that involve both types of data.

💡Open Source

Open Source in the context of AI refers to models and tools that are freely available for use, modification, and distribution. The video script emphasizes the importance of understanding both open source and paid models in generative AI, mentioning examples like Lama 2 and Mistral.

💡Paid Models

Paid Models are AI models that require payment for use, often providing additional features or support. The speaker in the video discusses the need to understand both open source and paid models, highlighting platforms like Open AI and Cloudy 2.

💡Fine-tuning

Fine-tuning in AI refers to the process of adjusting a pre-trained model to perform a specific task by training it on a smaller dataset. The video emphasizes the importance of fine-tuning custom data with LLMs as a crucial skill in generative AI engineering.

💡Frameworks

Frameworks in the context of AI are sets of tools and libraries that provide a structure for developing applications. The video script mentions frameworks like Lang chain and Llama Index, which are important for developing applications that utilize generative AI models.

💡Vector Databases

Vector Databases are databases designed to store and retrieve vector data, which is commonly used in AI for tasks like natural language processing. The speaker in the video discusses the importance of understanding vector databases like Chroma DB for working with text data in generative AI.

💡AWS Bedrock

AWS Bedrock is a service mentioned in the video that provides access to various AI models, including both open source and paid models. It allows for the integration of these models into business applications and emphasizes the role of cloud services in deploying AI models.

Highlights

Introduction to the importance of skill sets for becoming a generative AI engineer.

Discussion on the increasing demand for generative AI in data science and solving business use cases.

Emphasis on the necessity of understanding generative AI for those interested in working in this field.

Announcement of providing necessary materials and a playlist for learning generative AI.

Explanation of generative AI engineering and its prerequisites.

Description of the three types of generative AI models: large language models, large image models, and multimodal models.

Differentiation between large language models and their applications in text-related use cases.

Highlighting the competition among tech giants in developing accurate large language and image models.

Introduction to the concept of multimodal models that combine text and image for solving complex use cases.

Importance of understanding the main aim of generative AI models to generate new content based on context.

Categorization of generative AI into open source and paid models and the need to learn both.

Discussion on the use of open source models like Lama 2 and their potential for business use cases.

Mention of paid models and services like Open AI, Cloudy, and AI 21 lab for advanced functionalities.

Exploration of the role of cloud platforms in deploying and scaling generative AI models.

Introduction to AWS Bedrock and its comprehensive service for integrating various LLM models.

Recommendation to explore frameworks like Lang chain and Llama Index for developing generative AI applications.

Emphasis on the importance of practicing with different models and understanding their limitations.

Highlighting the significance of fine-tuning models with custom data for specific business needs.

Provision of a roadmap and resources for learning prerequisites in generative AI engineering.

Transcripts

hello all my name is Kish naak and

welcome to my YouTube channel so guys in

this particular video we are going to

discuss about all the important skill

sets that you may specifically require

in order to become a generative AI

engineer the reason why I'm making this

specific video right now most of my

students who have already transitioned

into the data science field who are

working as a data scientist they're

getting a lot of work to work in the

field of generative AI specifically with

respect to solving various use cases

with the help of large language models

and large image models so if you are

really interested in understanding about

generative AI work and you really want

to work in this field then this video

will definitely be for you and the most

important thing of this specific video

will be that I will be providing you all

the necessary materials and all the

entire playlist where I've created a lot

of videos and probably in the upcoming

one or couple of months you'll be seeing

a lot of end to end projects that will

be coming up uh I will try to use lot of

tools specifically that you actually

require to solve this business use cases

considering both open source llm models

and paid llm models and I'm already

doing that anyhow in the description

link I'll be providing you with all the

materials and all the playlist link now

let me quickly go ahead and start

explaining about generative AI

engineering and what are the skill sets

that are basically required whenever I

talk about generative AI engineering

right so I really want to talk about two

important thing one is the prerequisites

okay like if you really want to get

enter into this field what are the

prerequisites the second thing is that

what are the important skill sets Okay

so this entire video I will be talking

about this two important thing with

respect to prerequisite let me just

explain about the prerequisite in some

time okay so I will take this up in some

time but before this let me talk with

respect to skill sets Okay now and here

I'm just going to focus on generative AI

as I said prerequisite what are the

necessary things that you really need to

know so that in the interviews whatever

things are basically asked for machine

learning deep learning NLP that will be

part covered in entirely in this

prerequisite itself when I talk about

skill sets this is specifically related

to generative AI okay generative AI so

with respect to skill sets if I consider

this

okay understand one one one thing right

what exactly generative AI is generative

AI basically

means generative AI I basically means

here you are trying to work with those

kind of models right and specifically

when I talk with respect to generative

AI there are two types of models that

you will probably see right right now it

is three one is llm model large language

model the second one is large image

model and the third one is something

called as

multimodel okay multimodel if I probably

see with respect to categorization these

are the three important types with

respect to generative AI models okay so

one is llm large language model large

image model and multi model let me talk

about the differences between them large

language model basically you

specifically work with use cases that

are related to text and these models are

trained with huge amount of data okay

that usually trained with huge huge

amount of data nowadays there is a

fierce competition between big these big

giants who are able to create good large

language models or large image models

with the better accuracy so if you

probably see openi if you probably see

Google if you probably see Microsoft if

you probably see uh meta all these

companies are in a fierce competition to

probably make the best llm models or L

models right and over there they will be

using specific data sets at the end of

the day they are everybody's in

competition because once they achieve

that

accuracy and uh all the people will try

to use those specific models to solve

the business use cases if I talk about

multimodel it is a combination of text

and image so that basically mean that

model will be able to solve use cases

that involves both text and images like

example if I probably consider Google G

Pro and right now it is equivalent to

chat gp4 okay now our main aim is that

in generative AI main aim of the model

is to generate new content right based

on any context it'll be able to generate

new content so this is the main aim with

respect to generative AI itself now with

respect to skill sets now this is what

already have been probably researching

from 6 to 7 months I mean asking people

who are specifically working in this

specific field what kind of use cases

they are solving how they are solving

everything as such so with respect to

skill sets if I consider right and right

now this entire generative AI is

basically divided into two important

things okay one is Open

Source One is paid paid llm model

specifically if I llm or I can basically

say paid models Okay the reason why I'm

writing this because I'm going to

categorize guys based on this two

important information open source and

paid

models do you need to learn both of them

yes the answer is very simple yes you

really need to have the complete idea

I'll tell you right now the generative

AI field is somewhat like how in 2018

machine learning was right people were

exploring people were getting to

understand new things right how they can

actually Implement things how can they

can actually use this models to do the

deployment or solve complex problems how

we can actually make it scalable all

those things and right now platforms are

also coming so still you know companies

are using it they're seeing they're

seeing that how we can actually solve

the specific use cases so my always the

main main main thing that I really want

to talk about is that practice as much

as you can try to solve many use cases

as you can right tomorrow if you're

specifically working in a company if any

use cases come you really need to try

each and everything open source paid

everything right then you need to

understand that what are are the

downfalls what are the disadvantages

what are uh probably with respect to

open source and paid models which is the

best thing to probably go ahead with

right and automatically you'll be able

to understand okay this is the problem

that our company is facing this is what

is the perfect way to probably go ahead

with so that is the reason I've divided

this into two thing one is open source

llm models or L models paid models okay

and if I talk about open source

specifically you'll be hearing some

amazing companies like meta right uh

they have actually come with this Lama 2

right if I probably come up with this

Lama 2 Lama 2 model right now its

accuracy is good in some days probably L

3 will also come right uh not only Lama

2 like let's say consider paid one right

if I probably consider with respect to

paid one then you have open AI you have

cloudy right cloudy 2 you have one more

model is the a21 lab right sorry AI I

think it is a AI 21 lab right these are

they are also providing lot of

functionalities uh if I probably

consider one more uh Mistral right

mistol is also given as an open source

also and there is also a paid version of

it okay I guess both the combination is

there um recently I've started exploring

mrr also uh there's more open source

models like Falcon

right right now these all llm models if

I if I probably consider llm or L models

whatever things is there or here one

more example is Google gmany pro right

Google gy Pro right right now Google gy

pro is also given I'll not say it is an

open source it is also paid API right

now but it is also given in the free

version so with respect to Google ji Pro

here you'll be able to see that you'll

be able to use this and all right now

what is the differences between this

open source and paid models right now

you can use this entire open source

models to solve various business use

cases obviously you can use this to

solve business use cases you can also

use you can also use the paid models to

solve business use cases understand this

thing right but when we think in terms

of

deployment right we think about

scalability right so obviously we really

need to look at the cloud part when we

are specifically using this open source

model itself obviously the accuracy is

high over here accuracy is less that is

another thing but when we say that we

are taking this into the production

level at that point of time you'll be

seeing that we really need to be

dependent on the cloud right so one

thing that matters is with respect to

the cloud and this is also one of the

reason what business basically thinks

whether we should probably go with open

source and PR do we need to handle the

cloud right let's say I have a product I

want to use any of this models

functionalities right if I probably go

with open source I may have to use this

I need to finetune our data set and then

probably do the deployment right so the

cloud is one of the important factor in

the case of paid models they have their

own specific Cloud right personal cloud

let's say cloud A2 has some specific

Cloud open AI has their own apis itself

everything is basically provided with

respect to this where you can use these

apis along with to solve any business

use cases now this is one way now there

is also another way we can also be

dependent on

AWS or other Cloud platform like aure

let's say in AWS we have something

called as there is a service which is

called as AWS

Bedrock now AWS Bedrock I'll soon make a

tutorial a detailed tutorial I've

already started working on this and it's

superb guys it's superb right AWS

Bedrock what it does I I'll talk about

it okay now AWS Bedrock this is a

service what it does is that it has

almost each and every llm model stable

diffusion model LM model so one example

is all stable diffusion

right stable diffusion so what it does

is that this is basically specifically

used for image image for any use cases

that are related to image stable

diffusion okay now the best thing about

AWS Bedrock is that it has all the

functionalities with respect to this

open source or paid models it has each

and everything right I think it does not

have open AI other than that I think it

has almost everything right and this

actually

provides apis itself where you can solve

your business use cases you can perform

fine tuning right you can perform fine

tuning you can do each and everything

right probably use this specific API

directly into the clo so you don't have

to worry about the cloud part in this

right here also you don't have to worry

about the cloud part but right now let's

say if you want to use open AI then you

have a different API you if you want to

use Cloud 2 then you have to have a

different API right but what AWS Bedrock

has basically done is that it has

combined each and everything that is

probably available over here right so

when it is combining each and everything

over here in the form of apis that

basically means we can use this apis

directly over and fine tune do whatever

task we specifically require there uh

again when I probably show you the

detailed tutorial regarding AWS bedro

I'll show you how you can probably get

the API details how you have to probably

give the prompt and all what format it

is basically required everything will be

understood now what is super important

to understand you really need to have an

idea how to use this both open source

and paid models right one very important

way like one way is that you can go

ahead with AWS Bedrock but my suggestion

would be that try to explore different

different ways one way is specifically

by using hugging

face now using hugging face

you can call all these open source

models or paid models however you want

open source models right or other models

except openi you cannot call openi

because openi has a different way of

allog together doing them but when I

probably talk about Cloud 2 Mistral

right even Google gin Pro you cannot

right uh if you want to probably call

Lama 2 it has almost each and every

functionalities and what it does is that

it gives you set of libraries where you

can specifically apply embeddings also

uh you can probably use this all models

to probably solve any business use cases

not only that it also provides you Cloud

platform it also provides you space with

respect to Cloud where you can also find

tunate and here you probably have to go

pay as you go right pay as you go so the

more you use the services that much you

basically need to pay just calling this

particular thing models it is I think

almost similar like how we do with

respect to open source model so in short

the skill sets that you really need to

focus is open source two pay models paid

models you should know about the Su of

the information with respect to aw's

Cloud platform right Azure when I say

probably Azure over here you should know

that how you can use this AI services in

Azure itself right in Azure I think they

have functionalities that are related to

open and all okay since it is already

Microsoft and open AI are almost right

because Microsoft is investor in open AI

itself right

now very important thing that I have

actually mentioned over here along with

this I will talk about framework now

which Frameworks you really need to be

good at right till now guys one is open

AI framework you definitely need to know

how you can specifically use open a now

the main thing about the framework is

which I really want to mention is about

Lang chain and llama Index

right llama Index this two framework you

really need to be good at because this

two framework how it is basically

created if you really want to develop a

rag application I suggest always go with

ram andex if you want to create a

generic application and you want to

solve some problem statement you can use

Lang chain framework now what is so

amazing thing about this particular

framework because it provides you lot of

agent tools to perform different

different functionalities from data

injection to data transformation each

and everything the specific libraries

provides so because of this libraries

you'll be able to implement function uh

in a very much easy way and with Lang

chain and Lama index you can call both

paid and open source llm models that is

the best thing when I say llm models

please also consider L models or any

other models that I'm basically talking

about right both these Frameworks

provide you options to specifically call

this particular models itself so this is

also really good you should definitely

use Lang chain and Lama index one

more is nothing but chain late you

should also start exploring chain late

because this is also a framework that is

there now for practice sake you can use

framework like streamlit for the front

end right for the front end because this

will actually help you to create the UI

quickly and practice with respect to

this so all these things you

specifically need to learn with respect

to the skill sets now one very important

thing along with this you also need to

have knowledge with respect to Vector

databases there are lot of different

different Vector databases like chroma

DB right there is cassendra you can SP

specifically use cassendra you can use

data Stacks right to just get an

experience with working on this in the

production environment right and you

should try to see that how the specific

Vector databases work right once you're

able to understand the specific Vector

databases then understand developing any

application that will be related to text

you know you'll be able to convert those

text into some vectors and how you can

basically convert it they are different

different libraries which already

hugging face open AI already provide you

right so you can use this all

functionalities and important things

Vector databases is super important in

this specific framework now with respect

to task right Projects please try to

understand how to create projects how to

create projects right

using llm

models use Vector databases like pine

cone use techniques or Frameworks like

Lang chain right Lama

index Lama index see how you can use

specific agent tool see how you can use

databases see how you can actually do

the deployments each and everything

right so you need to create many

projects as possible once you create

many projects as possible tomorrow

whenever you go in any specific industry

you will have multiple option to

implement all these things right so this

is super important and one of the very

important task very very important task

and they may specifically ask you in the

interview is regarding finetuning of

your data fine tuning with your custom

data custom data using

llms this may be the very important

thing trust me I've written four star

over here but this is the most important

thing is nothing but fine-tuning with

your custom data

and you should know each and every

techniques over here let's say if you're

specifically using open AI models you

should know that if you're using llama

to you should know that if you're using

any other services you should know that

right through that way tomorrow whenever

you have a use cases where you really

need to work with fine tuning then this

will definitely help a lot okay so in

short I've given you the entire skill

sets and this is what I have done I've

created three amazing playlist one is

with respect to open AI one is with

respect to Lang chain I've covered

topics with respect to open AI Lang

chain and solve multiple use cases using

um different different models like

Google

gemin um then you also have llama 2

right I have actually covered all these

things right so all those playlist will

be given in the description of the

specific video right go ahead and enjoy

that video go ahead and check it out if

you're serious about it definitely every

everything is available I've have

provided that in the playlist all you

need to do is that go ahead and learn

things now let's go with respect to the

prerequisite this prerequisite I've

given the entire things you need to

learn before probably starting these all

things right if you already know that it

is very good so here I'll be showing a

road map which I have already shared

earlier in my videos road map to become

a generative AI so let's go ahead and

see the road map what all things you

should need to learn so guys just a

month back I had already created this

entire road map to become generative so

these are the prerequisites that you

really need to focus on and again all

the videos materials is given over here

one is Python programming language then

basics of machine learning and natural

language processing y NLP this this this

is there basics of deep learning right

how you should basically know that hown

works forward propagation backward

propagation optimizers activation

function Advanced NLP Concepts like RNN

lstm RNN GN by Dire hmrn and all so

these are some of the prerequisites you

can probably see Transformers is also

written over here all the video links

are basically given over here you can go

ahead and check it out so once you

probably go through this and then you

start developing more projects as

possible as you can the more you

practice the more scenarios you

definitely see the more you better you

become in the generative AI Feld so yes

this was it for my side I hope you like

this particular video I'll see you all

in the next video have a great day thank

you one all take care bye-bye