2024 Roadmap to Generative AI Mastery:

From Basics to Advanced Concepts

Here's a step-by-step guide designed for absolute beginners looking to

acquire skills in Generative AI.

The roadmap incorporates free learning resources for both technical and
tool-related skills.

Different Positions or Different Levels:

Developer Level 1 or Beginner Leve
Devoloper Level 2 or Senior Leve
Researcher Level

1 www.ineuron.ai
This Roadmap is broken into many sections:

Prerequisite

Fundamentals

Core Generative
Models

Developing
Applications

Powered by LLMs

Projects and Practical

Experience

Miscellaneous
Topics

Advice for
Productive Learning

FAQs

2 www.ineuron.ai
What is Generative AI?
Generative AI generates new data based on training samples. Generative
models can generate Image, Text, Audio, Videos etc. data as output.

So generative AI is a very huge topic,

Generative Image Model(GANs, Various Diffusions Models
Generative Language Model(LLMs)

When I refer to large language models, I mean natural language

processing. Since NLP forms the foundation of massive language-
generated models(LLMs).

Useful Learning Resource:

Gen AI Introduction Video In English-

https://www.youtube.com/watch?v=ajWheP8ZD70&t=3906
Gen AI Introduction Video In hindi-

https://www.youtube.com/watch?v=I-_a5hop_KU&t=3160
Gen AI Foundation Free Course -

https://bit.ly/3HcxVA
GenAI Introduction by Krish Naik-

https://www.youtube.com/watch?v=PoKwTzmrAts&t=1880s

3 www.ineuron.ai
Generative AI with Large Language Models

Prerequisite:

Programming Language:

Python is the most commonly used programming language for Data

Science, Machine learning and Ai domain.

Here are some reasons why:

Community Support: Active community of developers, researchers,

and practitioners in the machine learning and AI domains

Libraries and Frameworks: Many key libraries and frameworks for

generative AI, such as TensorFlow, PyTorch, and Keras, have Python

interfaces

Flexibility and Productivity: Python is known for its readability,

simplicity, and ease of use, making it an ideal language for rapid

prototyping and experimentation

Data Analysis and Visualization: Python is widely used in data

science, and it has excellent support for data manipulation and

visualisation libraries, such as NumPy, Pandas, and Matplotlib.

Topics to Learn-

Variables, Numbers, Strings

Lists, Dictionaries, Sets, Tuples

If condition, for loop

Functions, Lambda Functions

Modules (pip install)

Read, Write files

Exception handling

Classes, Objects

4 www.ineuron.ai
Useful Learning Resource:

Python Complete Playlist- https://bit.ly/4aK9SGR

SQL Database(Optional)
Deep learning projects often involve working with large volumes of
unstructured data, such as images, text, or audio. In many cases,
traditional SQL databases may not be the primary choice for storing
such unstructured data, and other types of data storage solutions
might be more appropriate
But for the machine learning project it is important to.
My Preference: MySql or Sqlite3

SQL Topics to Learn:

Basics of Relational Databases

Basic Queries: SELECT, WHERE LIKE, DISTINCT, BETWEEN, GROUP

BY, ORDER BY

Advanced Queries: CTE, Subqueries, Window Functions

Joins: Left, Right, Inner, Full

Useful Learning Resource

SQL Complete Playlist - https://bit.ly/3vBFMF7

5 www.ineuron.ai
NoSQL Database:

The need for a NoSQL database in a deep learning project depends on the
nature of your data and the specific requirements of your project.

Deep learning projects often involve working with large volumes of

unstructured data, such as images, text, or audio. In many cases, NoSQL
databases are used to store and manage such unstructured data
efficiently.

Reason by you should Nosql Database:

Scalability and Flexibility:

Variety of Data Types:
Real-time Data Ingestion:
Distributed Computing:
Schema-less Design:
My Preference: MongoDB or CassandraDB
Useful Learning Resource:
Fundamental Video of MongoDB-

https://www.youtube.com/watch?v=VIAcD6P_Etc

https://www.youtube.com/watch?v=KWoyJwqt22I&t=4498s

Fundamentals

Math and Statistics for Data Science(optional)

Reason why we need to learn it-

Math and statistics are fundamental for data science and AI as they draw meaningful
insights from complex datasets.

6 www.ineuron.ai
Topics to Learn in Statistics
Basics: Descriptive vs inferential statistics, continuous vs discrete data,
nominal vs ordinal data
Basic plots: Histograms, pie charts, bar charts, scatter plot etc.
Measures of central tendency: mean, median, mode
Measures of dispersion: variance, standard deviation
Probability basics
Distributions: Normal distribution
Correlation and covariance
Central limit theorem
Hypothesis testing: p value, confidence interval, type 1 vs type 2 error, Z
test, t test, ANOVA
Topics to Learn in Mathematics

Probability
Linear Algebra
Calculus
Useful Learning Resource:
Video of Statistics and Mathematics by Krish Naik- https://bit.ly/47u8qp1

Basic Deep Learning

Topics to Learn
Artificial Neural Networks
activation functions and Loss functions
Backpropagation, optimizers
Regularisation, Normalisation
Convolutional Neural Networks (CNNs)
Recurrent Neural Networks (RNNs)
Get hands-on experience with frameworks like TensorFlow or PyTorch.

Useful Learning Resource:

Deep Learning Community Session by krish Naik- https://bit.ly/48s3Jxs

7 www.ineuron.ai
Basics of Natural Language Processing
Topics to Lear
Text Preprocessing: Regex,Lowercasing,Tokenization,Removing
Punctuation,Removing Stop Words,Stemming,Lemmatization
Text Representation: Count vectorizer, TF-IDF, BOW,OHE
Text Classification: Naive Bayes
Fundamental library: Spacy & NLTK
Useful Learning Resource:
NLP Community Session by Krish Naik- https://bit.ly/3NYzkP7

Word Embedding Techniques:

Topics to Lear
Word2Vec
GloVe
ELMO
Fast Text
Useful Learning Resource:
Deep Learning Community Session by krish Naik- https://bit.ly/3NYzkP7

Advance NLP Concepts

Topics to Learn
Advance RNN like LSTM & GRU
Encoder decoder & Encoder decoder with Attention Mechanism
Transformer architecture: Self attention mechanism, key, query,
value(KQV), Layer Normalisation & Positional Encoding
BERT: Contextual embedding and mask language modelling
GPT: Autoregressive Modelling

8 www.ineuron.ai
An important concept needs to be learnt.
Transfer Learning: learned from past work and applied it to the

current challenge.
Fine-Tuning of Model: Fine-tuning refers to the process of taking a

pre-trained model and further training it on a domain specific task.

Different Sequence mapping: One to Many, Many to One, Many to Many

Useful Learning Resource:

NLP Community Session by Krish Naik- https://bit.ly/3NYzkP7

Core Generative Models:

Large Language Models(LLMs)
Topics to Learn
Milestone LLM Models
BERT: Bidirectional Encoder Representations from Transformers
(BERT) was developed by Google
GPT: GPT stands for "Generative Pre-trained Transformer".The model
was developed by OpenAI
XLM: Cross-lingual Language Model Pretraining by Guillaume
Lample, Alexis Conneau.
T5: The Text-to-Text Transfer Transformer It was created by Google
AI
Megatron: Megatron is a large, powerful transformer developed by
the Applied Deep Learning Research team at NVIDIA
M2M-100: multilingual encoder-decoder (seq-to-seq) model
researchers at Facebook

9 www.ineuron.ai
OpenAI LLM Models
GPT-4 and GPT-4 Turbo: A set of models that improve on GPT-3.5
and can understand as well as generate natural language or code
GPT-3.5: A set of models that improve on GPT-3 and can
understand as well as generate natural language or cod
DALL·E: A model that can generate and edit images given a natural
language prompt
TTS: A set of models that can convert text into natural sounding
spoken audio
Whisper: A model that can convert audio into text

Google AI LLM models:

PaLM2
Gemini-pro
Gemini -pro-vision

Meta AI LLM Models

LlaMA & LlaMA2

Open Source LLM Models

BLOOM
Llama 2
PaLM
Falcon
Claude
MPT-30B
Stablelm

Useful Learning Resource:

tt :// . /m
Hugging face Models Hub- h ps huggingface co odels
tt :// t m. . m/ /m
OpenAI Models- h ps pla for openai co docs odels
tt :// x v. / / 8 . 48 5
BERT Research Paper- h ps ar i org abs 1 10 0 0
tt :// t. y/48J w W
GPT Research Paper- h ps bi l H d

10 www.ineuron.ai
Prompt Engineering
Topics to Learn
Type of Prompting:
Zero shot prompting(Direct Prompting)
Few shot prompting
Chain-of-thoughts prompting
Prompt Creation: Length,context structure and specific instruction
Prompt Communities: PromptHero, FlowGPT, Snack Prompt
Useful Learning Resource:
Video link from Gen AI community session-https://bit.ly/3Saq0Ku

Developing Applications Powered

by LLMs-
Explore Generative Model APIs:
Topics to Learn
OpenAI API
Hugging Face API
Gemini API

Useful Learning Resource:

Link from GENAI community session course-

https://bit.ly/3vxuQbN
Link from Krish sir community session for GEMINI:

https://bit.ly/3NWiwZ5
Documentation Link-

https://ai.google.dev/docs

https://huggingface.co/docs

https://platform.openai.com/docs/introduction

11 www.ineuron.ai
Framework for Developing LLM application
Topics to Learn
LangCha
Chainli
LlamaIndex2
Useful Learning Resource
Link from GENAI community course-

https://bit.ly/3Saq0Ku

https://bit.ly/48smFM
Documentation Link:

https://docs.chainlit.io/get-started/overview

https://docs.llamaindex.ai/en/stable/

https://python.langchain.com/docs/get_started/introduction

Vector Databases
Topics to Learn
ChromaDB
Waviet
Pinecone
OpenAI Faiss
Useful Learning Resource
Link from GENAI community course-

https://bit.ly/3TNl934

https://bit.ly/3S9oOq
Documentation Link:

https://docs.trychroma.com/

https://docs.pinecone.io/docs/overview

https://weaviate.io/developers/weaviate

12 www.ineuron.ai
Tools and Framework for Web-Application
Topics to Learn
Streamlit
Gradio
FastAPI
Flask

Useful Learning Resource

Link from GENAI community course-

https://bit.ly/3TOVG9r

https://bit.ly/41UbbPo
Documentation Link: https://docs.streamlit.io/

https://fastapi.tiangolo.com/

https://flask.palletsprojects.com/en/3.0.x/

https://www.gradio.app/docs/interface

Deployment of LLM model

Topics to Learn
AWS
GCP
Azure
LangServ
HuggingFace Spaces
Useful Learning Resource
Link from GENAI community Course- https://bit.ly/41UbbPo

13 www.ineuron.ai
Few Advance Topics:
ChatGPT: Understanding of Chat Gpt Training and RLHF
(Reinforcement learning through human feedback) Concept
RAG : Retrieval-Augmented Generation (RAG) Systems
PEFT : Parametric efficient fine tuning
Adaptive Ranking: low rank adaptation(LoRa) and Quantized Low
Rank Adaptation(Qlora)
Evaluation of LLMs: Find evaluation metrics of results generated by
LLM

14 www.ineuron.ai
Projects and Practical Experience:
Hands-on Projects:
Work on small projects to apply what you've learned.

Experiment with different datasets and model architectures.

Kaggle Competitions and Open Source Contributions:

Participate in Kaggle competitions related to generative tasks.
Contribute to open-source projects in the generative AI field.
Useful Learning Resource
Link of iNeuron Internship Portal-

https://internship.ineuron.ai/
Link of iNeuron Job Portal-

https://jobs.ineuron.ai/

Miscellaneous Topics
Platform To Explore:
LIDA (Automatic Generation of Visualisations and Infographics)
Slides ( AI Presentation Maker )
Content Creation (Jasper, Copy.ai, Anyword)
Grammar checkers and rewording tools (Grammarly, Wordtune,
ProWritingAid)
Video creation (Descript, Wondershare Filmora, Runway)
Image generation (DALL·E 2, Midjourney)
Research (Genei, Aomni)
GANs:
Variational Autoencoders (VAEs)

Generative Adversarial Networks (GANs)

15 www.ineuron.ai
Stable Diffusion Models:
Deliberate,Realistic Vision etc.

Stable Diffusion Models:

High end performance GPUs (GCP, AWS, Azure), Data Crunch,
PaperSpace, Google colab and google colab pro, Kaggle instance etc.

Continuous Learning:
Keep up with the latest: news, trends, research paper and community
Useful Learning Resource
For latest Research and News -

https://www.marktechpost.com/

https://paperswithcode.com/

https://aimagazine.com/
Link of NeuronLAB:

https://neurolab.ineuron.ai/

Advice for Productive Learning:

Define specific, achievable learning goals
Consistent learning
Don't forgot to Implementing your learning
Experiment and Iterate
Constructive feedback

16 www.ineuron.ai
FAQs
Do I need a background in machine learning or deep
learning to start learning Generative AI?
While a basic understanding of machine learning concepts is
beneficial, some introductory courses in machine learning can help
bridge the knowledge gap.

Deep learning is a fundamental part of Generative AI, and a

background in it is highly recommended. Familiarity with neural
networks, backpropagation, RNNs and common deep learning
frameworks like TensorFlow or PyTorch is advantageous for
comprehending generative models.

How much mathematics knowledge is required for

Generative AI?
Mathematics is a key component of understanding the algorithms
behind generative models. Students inquire about the level of
mathematical knowledge needed, with a focus on linear algebra,
calculus, and probability. A foundational understanding of these
mathematical concepts is beneficial.

Can I start with Generative AI without prior experience

in AI or computer science?
Yes, you can start but a background in AI or computer science can
provide a smoother start, there are beginner-friendly resources
available to help newcomers build their skills. You can refer to this
video for more details.
Mastering Generative AI with OpenAI, LangChain, and
LlamaIndex Batch Launch: https://bit.ly/3NS1hI3
Course Link for Mastering Generative AI: https://bit.ly/3HeguzF

17 www.ineuron.ai