AI4Bharat

Venturing into the field of AI/ML

S V Venkata Sai Vishwesvar

Introduction

Are you new to the world of AI and machine learning? Does it all seem a bit daunting and
overwhelming? Unsure of where to begin or how to navigate this vast field? You’re not
alone. We've all faced those initial challenges and felt the uncertainty you might be
experiencing. But don't worry, every expert started as a beginner! Here’s a roadmap to ease
you into the fascinating realm of Artificial Intelligence. 😀

Getting started…
If you're starting with no coding or mathematical background, it's wise to spend at least a
couple of months mastering the basics of Python and Math. However, if you're already
comfortable with these, you can jump right in. Level 0 is designed for absolute beginners,
building from the ground up.

Level 0:

For an absolute beginner, I highly recommend Code with Mosh’s 6-hour Python tutorial on
YouTube. This comprehensive lecture covers everything you need to take your first step
into coding. Mosh is an exceptionally patient and passionate teacher who explains
concepts clearly and builds your understanding step by step, making it an ideal starting
point for those new to programming.

Link : https://www.youtube.com/watch?v=_uQrJ0TkZlc

1
Level 1:

For the next couple of weeks, I recommend brushing up on your basic calculus skills: limits,
derivatives, continuity, differentiability, basic integration, partial derivatives, and high
school linear algebra. Mathematics is the foundation of AI, and having a solid grasp of these
concepts is essential. Understanding the math behind AI algorithms will be immensely
helpful when you need to dive into the details of the models you’ll be deploying.

While these topics are enough to get you started, you’ll need to continuously upgrade your
math skills, particularly in calculus and linear algebra. To help with this, I'm linking two
excellent courses that I personally found invaluable:

1. Multivariable Calculus:

https://www.youtube.com/watch?v=XzaeYnZdK5o&list=PLtKWB-wrvn4nA2h8TFxzWL
2zy8O9th_fy

Watching just the first 10 videos of this playlist should cover most of the calculus
you'll encounter initially. However, I always advocate for learning more math as a
general rule of thumb. If you're eager to dive deeper and expand your
understanding, it would be an excellent investment in your AI journey. The more
math you know, the better equipped you'll be to tackle complex problems down the
road.

2. Linear Algebra by Prof. Gilbert Strang – A course that needs no introduction.

https://ocw.mit.edu/courses/18-06-linear-algebra-spring-2010/

Linear algebra is the backbone of AI, powering everything from data

transformations to the operations within neural networks. It allows us to efficiently
represent and manipulate large datasets and enables algorithms to learn by
optimizing complex functions. Without linear algebra, the advanced computations
that drive AI would be impossible.

2
 I’m not suggesting you complete this entire course in one sitting. It’s a true treasure
trove of knowledge, best appreciated and absorbed gradually. My sincere advice is
to revisit this course regularly, taking the time to fully grasp each concept.

______________________________________________________________________________________________

Level 2:

Now that you have laid down the foundations, it is time to move on to “actual ML” 😆
Prof Andrew Ng’s Machine learning specialization on CourseEra is one of the most popular
courses out there .This is a fantastic starting point for anyone new to the field. It offers
clear explanations of key concepts like supervised and unsupervised learning, as well as
practical tips for implementing algorithms. The hands-on assignments will help you build a
strong foundation and gain valuable experience in machine learning.

https://www.coursera.org/specializations/machine-learning-introduction

At this stage, it's also a good idea to start familiarizing yourself with key Python libraries like
NumPy, Pandas, and Matplotlib. These tools are essential for data science and will
significantly enhance your ability to work with data. To refine your skills and learn these
concepts, I've found the mini-courses on Kaggle to be incredibly useful and recommend
them for a practical, hands-on approach.

https://www.kaggle.com/learn/pandas

Learn Data Visualization Tutorials | Kaggle

https://www.kaggle.com/learn/feature-engineering

3
A slight detour…

Now that you've encountered Kaggle, let me give you a brief overview of what this platform
is all about. Kaggle is the ultimate destination for everything related to data science,
offering a range of resources from toy datasets to large-scale hackathons. It’s where a lot of
the magic in the data science community happens. Expect to visit this site frequently
throughout your machine learning journey, as it will be a valuable resource and hub for
your learning and growth.

Kaggle is a melting pot of talent from all corners of the data science world. You'll find
everyone from enthusiastic beginners to seasoned veterans working at leading
organizations like NVIDIA, Google DeepMind, and Hugging Face. These experts are actively
involved in the community, participating in and often dominating hackathons 😆. They
offer invaluable advice to newcomers, and contribute to the platform’s vibrant atmosphere.
Take full advantage of their wealth of knowledge and experience to support and enhance
your machine learning journey.

Being on Kaggle, or "Kaggling," has become a rite of passage for every machine learning
practitioner. It's a platform that most ML professionals have engaged with, and many
continue to do so throughout their careers. Whether you're diving into beginner-friendly
hackathons like “Titanic” or tackling complex real-world competitions, Kaggle is your go-to
destination. Welcome aboard, and get ready for an exciting journey in the world of data
science! 🥳
As a beginner, being on Kaggle offers you a unique opportunity to accelerate your learning
in a practical way. Beyond just accessing datasets and participating in competitions, Kaggle
also provides a rich repository of kernels (code notebooks) where you can explore how
others approach problems, see different coding techniques in action, and even collaborate
on projects. This hands-on exposure will deepen your understanding and help you develop
problem-solving skills by learning from the community’s diverse approaches. Embrace this
treasure trove of resources, and you'll find yourself growing faster and more confidently in
your machine learning journey.

4
Level 3:

Now we arrive at the juicy part of the field—drum rolls, please! Welcome to the world of
deep learning, where the real magic happens today. Deep learning has revolutionized
numerous fields, from Natural Language Processing (NLP) and Computer Vision (CV) to
Automatic Speech Recognition (ASR). It’s the driving force behind remarkable
advancements like ChatGPT, which relies on sophisticated neural networks to understand
and generate human-like text.

Deep learning models are designed to mimic the way the human brain processes
information, allowing them to learn and make decisions from vast amounts of data. This
has led to groundbreaking applications and innovations that are shaping our world.

To dive into deep learning yourself, I highly recommend Prof. Andrew Ng's Deep Learning
Specialization on Coursera. It covers essential topics like neural networks and their training,
techniques for improving network performance, structuring machine learning projects,
convolutional neural networks for image tasks, and sequence models for NLP and speech
recognition.

https://www.coursera.org/learn/neural-networks-deep-learning

Side note:

The practical assignments in the course are particularly valuable, offering hands-on
experience that solidifies your understanding.

Here’s a bonus tip: Once you’ve completed the programming labs, challenge yourself by
erasing the provided template code and coding from scratch. This will elevate your
conceptual clarity to a gigachad level! 🤭

5
Level 3.5:

At this stage, you have enough expertise to step into the real world and start building and
deploying models. However, now is the time to dive deeper into a more math-intensive
course, because, as I’ve emphasized throughout, math is the bedrock of AI. Concepts that
were cutting-edge just two years ago are now fundamental to the field.

For those who are truly dedicated to mastering AI, here are two comprehensive courses
that will solidify your understanding of deep learning and its core concepts:

(by my Prof, yay! Hi Professor! 😀)

1. Introduction to Deep Learning (Prof. Mitesh Khapra) This course provides a
thorough, holistic understanding of the field, ensuring that your fundamentals are
rock-solid. What I personally appreciated is how deeply it explores the mathematical
intricacies of key topics, such as backpropagation, optimization strategies, CNNs,
and backpropagation through time.
Bonus tip: Try to solve the assignments on the course page—they’re incredibly
insightful and well worth your time.

Link: http://www.cse.iitm.ac.in/~miteshk/CS6910.html

2. Introduction to Large Language Models (Prof. Mitesh Khapra) Given that

generative AI is heavily reliant on LLMs, this course is a must for those looking to
challenge themselves and elevate their understanding. Prof. Khapra covers the
latest advancements in the field, including topics like encoders, decoders,
tokenization, attention mechanisms, embeddings, and more.

Link: https://www.youtube.com/watch?v=qrsNX1Rwle0&list=PLZ2ps__7DhBbaMNZoyW2Hizl8DG6ikkjo&index=1

(Please note, this course is still in progress. So make sure to check back for updates
regularly 😀)
Mastering these courses will significantly enhance your understanding and give you a
strong foundation in AI.

6
Level 4:

Now that you're familiar with the fundamentals of deep learning, it's time to learn a deep
learning framework to start building models for specific tasks. There are several
frameworks available, like TensorFlow, PyTorch, and JAX, but I recommend starting with
PyTorch. PyTorch is not only flexible and easy to use, but it's also the framework of choice
for many top companies like Facebook, Microsoft, and Tesla. Additionally, PyTorch's strong
community support ensures that you’ll have plenty of resources to help you along the way.

One of the best resources for learning PyTorch is ZTM’s PyTorch course, taught by Mr.
Daniel Bourke. I personally took this course and enjoyed every moment of it. Daniel starts
from the absolute basics and introduces the concepts beautifully. He doesn’t just
spoon-feed information but encourages you to stand on your own two legs, teaching you
how to solve problems independently. This is the point in your journey where you begin to
grow your wings and step out into the real world.

Daniel is an amazing teacher whose infectious passion is bound to rub off on you. The
course also includes extensive online reference material with plenty of exercises to practice
and strengthen your fundamentals. I loved every second of this course, and I’m confident
you will too!

Link: www.udemy.com/course/pytorch-for-deep-learning/

Please take the time to follow along with the course religiously. Make sure to solve all the
assignments provided, as they are crucial for reinforcing your understanding. Take the time
to relish each concept, appreciating its inherent beauty and the powerful insights it offers.
Deep learning is a journey, and immersing yourself fully in each step will help you build a
strong foundation for everything that lies ahead.

7
Some General Pointers:

Whew, folks! What a journey we've been on together. By now, you should feel confident
enough to step into the world of AI and ML without fear.

However, before you embark on this exciting adventure, there are a couple of things I'd like
to emphasize. First, regularly follow the veterans of the field. No post of mine would be
complete without mentioning my personal hero, Andrej Karpathy (All hail the 🐐!). He has
some of the best AI lectures available on his YouTube channel, and I cannot stress enough
how useful and insightful they are. It’s an absolute gold mine to have one of the leading
experts in the field share such high-quality tutorials. Please take advantage of this
material—it’s nothing short of a treasure. 🥰
Link: https://www.youtube.com/@AndrejKarpathy

And here’s an exciting update: Andrej Karpathy is launching courses from his new venture,
Eureka Labs. He gave us a sneak peek of the LLM101 course a few months ago, and we’re
all eagerly waiting for it to drop, Andrej! 😍
Of course, there are other legends in the field, like Yann LeCun, Yoshua Bengio, and Ian
Goodfellow. The entire AI revolution has been shaped by their groundbreaking
contributions, and the world is deeply grateful for their work. I encourage you to watch
their interviews, read their papers, and learn from their insights—they’ve laid the
foundation for everything we do in AI today. 🙏
I also encourage you to read as many papers as you can to keep yourself abreast of the
field. Tinker with models, run inferences,build pipelines and keep learning—your curiosity
will drive you farther than you can imagine.

Once you start reading papers, take it a step further by replicating them. Choose a paper
that interests you, understand its methodology, code it out, and see the results for
yourself—then repeat. This practice is one of the best ways to gain a deep and robust

8
understanding of the field as a student. It’s through this hands-on approach that you’ll truly
master the concepts and techniques in AI.

That’s it, folks! What an incredible journey we’ve been on together. Wishing you all the
success in your AI endeavors, and may the algorithms always be in your favor! The world is
your dataset, and I can't wait to see the innovative ways YOU 🫵 will shape the future of AI.
Welcome aboard—we’ve got this! ⭐✨
Before we part ways, I’ll be leaving a collection of resources like books, playlists, lectures
and documentation down below. This is by no means exhaustive—nothing can be! But it’s a
good place to start exploring… 😎
Miscellaneous resources:

1) Scikit Learn docs: https://scikit-learn.org/stable/

2) PyTorch docs: https://pytorch.org/docs/stable/

3) Hugging face courses: https://huggingface.co/learn

(PS: I have not mentioned this above. But this is absolute gold)

Some Books:
1) Deep Learning: Ian Goodfellow, Yoshua Bengio, Aaron Courville

2) Learning Deep Learning: Magnus Ekman

And the best resource of all,

Your curiosity. 🔥 ❤️‍🔥

Wishing you all the best for your career. Cheers!

Signing off…
Venkat Shadeslayer 🥳🥳

9
10