Many questions are added regularly; please feel free to check back here from time to time.
At the outset, let us assume we have no prior idea about this subject — starting from absolute zero.
Building an ‘intelligent system’ currently requires a few foundational components. The story of bridging humans and machines — how to transfer human-like ‘thinking and reasoning’ to machines — is told step by step in our introductory books, notebooks, and videos. Following that, separate content explores the interactions between hardware machines and software, starting with basic concepts somewhat akin to Arduino. Interestingly, the very first stepping stone of this journey is ‘machine learning’. Here are answers to some initial questions.

1. I know nothing at all — where do I begin?
or,
2. I am a student of BBA / Polytechnic / Medicine. I have no programming experience. How do I get started?
-
Machine learning is for everyone. It is entirely possible to begin using common sense. In fact, people outside the computer science field are using it the most: Business Administration, Linguistics, Healthcare, Law, Human Resources, and many more. If someone wants to build English-to-Bangla translation or Text-to-Speech using machine learning, who will do it? A linguist. Who will determine which employees might leave your company in the coming months using machine learning? An HR professional.
-
Treat machine learning as a general tool, much like basic arithmetic. This tool is now being utilized across almost every discipline. The earlier you learn it, the further ahead you will be in your own field. One thing is evident: those without a computer science background often feel deeply intimidated by it. In reality, this subject is relevant and accessible to people from all academic backgrounds.
-
Machine learning does not involve the kind of rigid traditional programming that people imagine. What is there can easily be understood through common sense and basic logic. Furthermore, useful code snippets for various tasks are readily available on platforms like Stack Overflow. Once again: do not be afraid at all!

The reading progression for starting machine learning: Books 1 through 4.
3. How do I get started? Where can I find the books?
Four books for hands-on learning. The reading progression for starting machine learning is from 1 to 4. The final book in the image covers deep learning (see below). After finishing the two machine learning books, you can move on to the deep learning book. Following that are two books covering an application of deep learning: Natural Language Processing. All these books are freely available online. To explore all the books together in one place, see the Rokomari link.
| Book Title | Online Link | Print Book |
|---|---|---|
| Hatekolome Machine Learning (2nd Edition) | Gitbook | Rokomari |
| The Python Part Added in 2nd Edition | Gitbook | Rokomari |
| ‘Python Machine Learning from Zero’ (2nd Edition) | Gitbook | Rokomari |
| Hatekolome Python Deep Learning | Gitbook | Rokomari |
| Hatekolome Bangla Natural Language Processing | Online Book | Rokomari |
| A Humanistic State through Artificial Intelligence (Online) | Online | Interface |
| In Search of Excellence: Bangladesh; A Humanistic State through AI | Online Book | Rokomari |
| Data Analytics for Non-Programmers | Video | Book |
| Bundle Link for 5 Books Together | Coming Soon | Rokomari |
| Buying from Neelkhet | Print books at Neelkhet: Haq, Manik Library & others | Phone: 01735742908, 01820157181 |
4. Why should I read ‘Hatekolome Machine Learning’ (Second Edition)?
For starting machine learning, books are the best medium. These books are written in a hands-on ‘step-by-step’ style. No prior programming experience is required to begin any of them. Link to Book 1: Here (the second edition is sufficient). If you do not wish to buy the book, you can read the complete first edition online for free. Link: Here. The accompanying video series for this book is available as a YouTube playlist here.
The first edition was built entirely in the ‘R programming’ environment. It has been replaced by the second edition, which added Python. Those who purchased the first edition do not need to buy the second edition. Those who have the second edition do not need the first edition.
5. Who is ‘Python Machine Learning from Zero’ (Second Edition) for?
For those who want to learn machine learning exclusively using Python, but do not know Python yet, here is another book: Python Machine Learning from Zero. Link to the print book: here, or to read it online: here. The video series for this book is available as a YouTube playlist here.
6. How should I use ‘Hatekolome Python Deep Learning’ (Online and Hardcover Editions)?
For those who want to learn deep learning, the learning path begins with the Python book Python Machine Learning from Zero followed by Hatekolome Python Deep Learning. You can find the print edition here. You should not start the second book without completing the preceding one. The book is available online.
7. Are the books available in physical bookstores?
or,
8. Which bookstores in Neelkhet carry them?
(e) Hard copies are available at Rokomari, Adarsha, Baatighar, and several other libraries and bookstores. In Neelkhet, Dhaka, they can be found at Manik Library (01735742908) and Haq Library (01820157181 / 01511157181 / 01743516344).
9. I prefer learning through video — where are the YouTube links?
(f) You will find a 100+ video series on machine learning on YouTube and Facebook organized as playlists.
- YouTube Playlists: and Hands-on 1st Edition and Hands-on 2nd Edition; many more playlists are on the channel.
- YouTube Channel:
- Facebook Playlist: — look for the “Hatekolome Machine Learning (2nd Edition) Video Series”.
10. Between the green-yellow, red, and white books — which one should I start with, and when is each appropriate?
- Black book = Out of print, was the 1st edition of the red book (not needed)
- Red book = Hatekolome Machine Learning, 2nd Edition (R and Python) (no prior programming language knowledge required; everything you need to learn is inside the book)
- White book = Python Machine Learning from Zero (Python only)
- Green-yellow book = Hatekolome Python Deep Learning (Hardcover)
10.1 In what sequence should I read the books?
(Machine Learning / NLP Books)
1. ├── Red book = Hatekolome Machine Learning (2nd Edition)
-. | or
2. ├── Black book = Out of print; was the 1st edition of the red book
3. ├── White-yellow book = Python Machine Learning from Zero
4. ├── Green-yellow book = Hatekolome Python Deep Learning # For advanced users
5. └── Blue-white book = Hatekolome Bangla Natural Language Processing # For those working with language
(Data Analytics Books)
7. Upcoming
(Non-technical Books, For Everyone)
8. ├── In Search of Excellence: Bangladesh; A Humanistic State through Artificial Intelligence

The red book is designed for those who know nothing about machine learning and want to begin, have no programming background, yet want to understand machine learning intuitively. This book begins with the R programming environment, and solves the same project using Python as well.
Table of contents: http://bit.ly/ml-ho-2nd
The white book is aimed at the same audience, with the addition that it is for anyone who wishes to start exclusively with Python. This book is covered solely using Python.
Table of contents: http://bit.ly/ml-sk-py
Starting machine learning in the R environment is very straightforward. You do not need any prior programming experience to complete both books. Both books provide ample background context so everything is understood right then and there. Across the two books are two world-renowned, distinct real-world projects.
Afterward, if you wish to learn deep learning, you can use the Hatekolome Python Deep Learning (Hardcover) book.
Before buying the print books, you can read them on Gitbook. The print editions contain about 20% updated content. Read them online first, then decide whether to purchase.
10.2 What computer hardware configuration do I need to run the exercises in all your books?
To do the exercises across all my books (machine learning and deep learning) for data science, you do not need to buy any high-end PC. Almost all of my own work runs on a $150 Chromebook. A Raspberry Pi is even better. There is no need to buy an expensive computer to start learning data science at this stage. All you need is a web browser and an ordinary internet connection. All the work happens inside the browser.
For your convenience, I have hosted everything on the cloud and GitHub. Alongside this, I have provided Google Colab links, which directly download large datasets (Git LFS) into Google’s virtual machines in an instant. Those virtual machines have the necessary GPUs (if required). Those without a PC can even connect a smartphone (with a $7 adapter) to a monitor and carry out their work. Learning is what matters, not hardware.
11. Why should I learn machine learning?
Answer: The straightforward answer is: not everyone needs to learn it. First, ask yourself: what task am I currently doing using conventional methods that either cannot be done or takes too long, which machine learning could accomplish rapidly? Define your problem first, then see whether it is genuinely a machine learning problem.
Treat machine learning as another tool in your toolbox. One-breath reading link.
12. How do I know if a problem can be solved with machine learning?
Answer: Typically, machine learning answers stem from historical data in our possession. Sometimes the answer is a binary yes or no — a classification problem. Other times it is a continuous numerical output, such as: how many months until our return on investment materializes, or when might a customer churn? We call these ‘regression’ problems.
Alternatively, we might segment customers into cluster groups based on their behavioral patterns to provision new services. This represents an ‘unsupervised’ learning problem. One-breath reading link.
13. How long does it take to grasp the beginnings of machine learning?
Answer: Two days. However, to build intuitive understanding, I encourage you to look at “Project Titanic”. For prospective data scientists, this project is the gateway to machine learning. Seeing how a real-world problem is solved end-to-end makes everything clear. Titanic Link. Interestingly, the book below was written by solving this exact problem. Rokomari Link.
13.1 Do I need to know programming, statistics, and math at the beginning?
Short answer: No. But if you never step onto the road, how will you learn the path? My philosophy is simple: step onto the road. If we never lose our way, how will we know whether we are on the right track? Over the past five years of answering thousands of inbox questions, one pattern has stood out sharply: a large percentage of people ask the exact same question over a prolonged period. It is as if the person who asked the question is still asking it months later. This indicates they are caught in a state of paralysis and confusion. My unequivocal answer is: no, you do not need them at the very start. A major part of data science is ‘domain knowledge’. In healthcare and agriculture, where data science is making massive inroads, did practitioners spend years mastering pure programming, statistics, and mathematics before starting?
Although problem solving and programming are related, anyone who understands how a human solves a problem can begin machine learning. What you need at the beginning is an understanding of how to represent data in the context of the problem. And the answer depends on whom you ask: programmers will tell you one thing, while people who work with real-world data have an entirely different perspective. Truthfully, there is ongoing debate about how much programming skill is genuinely required in the machine learning ecosystem. And why wouldn’t there be? Machine learning was born precisely to break free from traditional rule-based programming.
What happens in conventional programming? Suppose you are programming to automate a process. You tell a robot to travel from point A to point B. That means you must specify every detail of the route: walk 10 steps forward at 90 degrees, then 20 steps to the right, then 15 steps at 170 degrees, and you will reach point B. If it encounters an obstacle anywhere along the way, it halts with an exit code. Now if you tell it to go to point C, all previous instructions are invalid.
What does machine learning do instead?
We show the machine many different routes for going from one point to another. When it runs through thousands of routes, it discovers a single underlying pattern: as it moves from the starting point toward the destination, the remaining distance decreases. Therefore, it chooses the path that minimizes distance. Afterwards, we test it on an unfamiliar road to see whether it can navigate successfully. Of course it can — because it understands the pattern. It is the difference between giving someone a fish and teaching them how to fish, so they never depend on others.
Can this be understood without programming? Absolutely. There is no need to worry about what critics say right now. I say: no, you do not need programming at the very beginning. What you need is a tool like Microsoft Excel, or the ‘R’ package — which takes just one day to learn. I have written a dedicated post on this.
Many advise mastering complete mathematics and programming before setting out on this journey. That is entirely mistaken. In reality, by the time someone finishes trying to learn everything exhaustively, their enthusiasm for the core subject has evaporated. Everyone agrees that learning what you need, when you need it, is the most effective approach. Some will tell you to learn the basics first before touching ML. It is unfair to claim you cannot succeed without doing that first. I myself only delved into higher-degree polynomials (the deep underlying math) when trying to understand a specific problem — never beforehand. I dug in, asked myself how much understanding was necessary to get the job done, and stopped there.
When you step onto the road, you learn as you move forward. At nearly fifty years old, I am continually learning new things; I certainly do not feel the need to ask anyone whether I should learn this or that. For instance, when list operations are needed in machine learning, and a list-of-lists operation is required alongside the basics, I learn the zip operation under the pressure of necessity. If you want to prepare, simply understanding variables, lists, arrays, loops, if-else statements, and dictionaries is more than enough. If you know how to interact with the operating system — reading, writing, and saving files — half the battle is already won!
Begin machine learning: wonderful tools exist that allow you to explore data without knowing programming — such as NumPy and pandas. You do not need deep programming knowledge. But as you dig deeper, you will notice where these tools reach their limits. A quick Google search will reveal that writing a simple loop or defining a helper function avoids having to write the same code repeatedly.
Hit the road. Once you learn to navigate by getting lost, nobody will be able to sell you hollow theories. Learning is an experience; I cannot explain riding a bicycle through a thousand lectures. You will fall many times; nobody learns without scraping their knees. Throw yourself into the challenge. If you cannot learn it, perhaps this path is not for you. Machine learning exists precisely so we do not have to write manual code. We can talk about that another day.
14. Which environment makes it easiest to learn machine learning?
Answer: Did you know that machine learning can even be learned using Microsoft Excel? Today, machine learning work takes place across several global environments. The “R” programming environment emerged from statistical computing — meaning “R” was built specifically around data. If you ask me, learning machine learning with “R” is far easier, especially at the beginning. It makes gaining intuition into the data much simpler. Alongside it, Python is universally used in machine learning. At the start, it is best not to fret over the programming environment. Learning becomes straightforward once you understand the philosophy behind machine learning. Link: Why R makes learning easy.
15. Which is better to start with: “R” or “Python”?
Answer: The responsibility of teaching is mine — over the span of one year. Leave it in my hands. You will be prepared using whichever tool is needed, where it is needed — largely language-independent. First, we will ground your concepts firmly. Detailed answers here: Philosophy-backed learning and Philosophy and how to begin.
15.1 If starting with Python, what are the absolute essentials I need to know?
You can start without knowing Python at all. However, here is a guideline for basic Python (which can be learned in two days):
(You can learn Python while learning machine learning — picking up tools as needed. For example, when list operations are required in ML, learn zip under operational necessity. If you wish to learn beforehand: basic variables, lists, arrays, looping, if-else, dictionaries, and file operations with the OS — reading, writing, and saving — cover half the work!)
Item by item, basic Python:
a. Variable assignment, numbers
b. Function calls, creating functions
c. Boolean operations, comparison operators, conditionals — if, elif, and else
d. Lists, indexing, slicing, list functions, working with tuples, and dictionaries
e. Various imports, interacting with the file system — how to read, write, and save files
See the link: Basic Python.
16. Can I grasp the core concepts in a single sitting?
Answer: Yes, you can. Watch 20+ short videos in 30 minutes. Check the playlists on Facebook and YouTube.
17. I did not study Computer Science — will I be able to do this?
Answer: I know many people working in the healthcare sector. Remarkable work is happening in healthcare with machine learning: novel drug formulas, early disease detection, and neuroscience research led by medical doctors. What you need for machine learning is ‘domain knowledge’ in your own field. For the rest of the code, four books in Bangla are right there for you.
18. What did you learn yourself?
Answer: In my own case, I learned both at the beginning — because each excels in different domains. ‘R’ is superior for starting and understanding data; Python is superior at production scale. Use whichever fits the task: a rickshaw for short distances, a motorbike for long journeys. You have to know what is needed where.
We live in an age of optimization. You may need to learn additional tools as requirements arise. If you let embarrassment hold you back, you will learn nothing. Even approaching fifty years of age, I must learn new things every single day. If you don’t keep learning, anyone can fall behind. Right now, the scikit-learn library fascinates me. It is so enjoyable that encountering a problem genuinely lifts my spirits.
19. I want to learn machine learning exclusively using ‘Python’ — is that possible?
Answer: It is possible, but given the time it takes, there is a high likelihood you might lose motivation. You can test this yourself: the complete exercise is implemented in Python at the link below. Even for those who already know Python, learning through R first is much easier — my experience confirms this. See the Python notebook.
The underlying concepts are identical everywhere. Yet, I repeat: intuition is formed much better with ‘R’. As mentioned before, for non-computer graduates or those wanting a hands-on, ground-up understanding, ‘R’ is on another level.
Believing that having a CS degree and knowing Python automatically makes you understand machine learning is a misconception. Python is a capable framework for many things; however, that does not mean ‘machine learning’ and ‘Python’ are synonymous. You will see that clearly as you advance.
20. What single skillset is most essential for learning machine learning?
Answer: Immersing yourself in data. To identify underlying patterns within data, you have to live with data — all day long.
21. How do you find joy in this?
Answer: Whenever I get free time, I apply machine learning to solve challenges in my job or at home. I try to plot everything to see whether hidden patterns emerge. You could say it is like using a cannon to shoot a mosquito. Perhaps it starts that way; but through making errors, I gain deep intuition.
Everyone looks for a central place where everything comes together — hence this site.
22. What platforms are available for learning machine learning in Bangla?
Answer: You can begin in three ways:
a. For those who prefer books and hand-notes: three books have been written and are available online. They are crafted so you do not even need video, though watching videos speeds up learning. Links are provided above.
b. For those starting with video: several playlists are available. On our Facebook page, click the “Videos” tab to find playlists organized by topic.
c. On YouTube, there is a dedicated channel with organized playlists. Clicking “Learn More” on our Facebook page takes you directly to the YouTube links.
23. Could you share the YouTube link?
YouTube @aiwithr — from there you can navigate directly to the playlists.
24. What is the current job market situation?
Management teams and founders of several e-commerce platforms and service industries frequently ask me to help them find “good” data scientists. However, I have been able to refer very few people — because our data scientists and industry use-cases do not yet align at that level of execution. This is not their fault; it is a market gap. In a nascent industry, there is a divergence between company expectations and candidate readiness. This gap will not last forever.
Our data industry is maturing; however, due to a shortage of well-labeled enterprise data, outcomes take time to materialize. On a promising note, professionals from other industries who consult with me on applying ML to their sector’s specific problems learn remarkably well. Ultimately, those who can define their problem clearly discover solutions the fastest.
Summary: Wherever you are, in whichever industry you work, try solving a real problem using machine learning. Challenge yourself. You will learn at lightning speed. Then you will be well prepared to transition into companies that focus purely on data science.
25. There are hardly any jobs in Bangladesh — what should I do?
Many beginners wonder: what is the point of learning this? Will I land a job? There are barely any jobs in the Bangladeshi market! So why should we learn machine learning? The few jobs that exist demand extensive experience, which has become a major source of frustration. I personally hear from many who feel disheartened. What is your advice for them?
An excellent observation. Thank you. You are entirely correct: everyone naturally wants a job after acquiring a skill. That is completely expected. Many ask me this, and I do not view it as unusual.
I tell everyone: over the course of our lives, we must develop multiple skillsets. Gone are the days when you could enter a single job and stay there for life. The times are far more demanding. I hold over 70 certificates across diverse disciplines. I realize now that certificates and degrees hold little weight when an employer sees that a candidate cannot deliver what is needed.
A company hires someone to solve a specific problem. Degrees and certificates serve merely as preliminary screening filters when the market is flooded with applicants. However, companies increasingly bypass rigid degree requirements when assessing whether someone can solve the task at hand.
In my own career, I learned whatever was needed, whenever it was needed. Over 28 years of professional life, I have reskilled myself 7 times. Talented colleagues and programmers I work with continuously update their skillsets to stay ahead of the game. Nobody wants to become sidelined or redundant. Retirement is a different matter.
26. What is your career advice in a nutshell?
-
When someone treats mastering a skillset as a vital personal mission, the outcome is extraordinary. Put everything into mastering the specific skill, not merely hunting for a job. Learn the necessary tools and stay updated. When you do, employers look at your GitHub profile rather than social media profiles. Companies will speak with you directly about their specific challenges. If you can outline solutions to their problems, landing the role becomes straightforward. Active participation and helping others in online forums also opens doors to unexpected opportunities.
-
In past generations, people could get by with a narrow set of skills throughout their careers. That no longer works. You must acquire multiple skillsets over a lifetime. Along the way, unlearning outdated methods and embracing new skillsets — what we call reskilling — is mandatory.
Finally, the market is emerging. Many people I know are receiving data science offers from Europe and Japan. The market is developing. Think globally, not just locally. While working in your current role, or from wherever you are right now, start learning machine learning. “Who knows when lightning could strike?”
27. How do I learn ‘Deep Learning’, and how should I proceed?
I wrote an extensive guide on this: read it here. I also explain why you should learn deep learning.
28. Where can I learn hands-on Bangla Natural Language Processing?
The book Hatekolome Deep Learning contains four chapters on Bangla Natural Language Processing. Another dedicated book is in progress: check nlp_bangla on GitHub.
29. A request: will you write a book on Machine Learning Mathematics and Algorithms?
My perspective is different: I focus on practical, applied engineering. Academic researchers develop mathematical foundations and algorithms because their domain is theoretical development.
In contrast, my mission is applying existing global technologies effectively to solve human problems. That does not mean I omit math entirely. But if everyone walked the same road, where would we be?
30. How can I contact you if I run into problems?
Answer: While I am passionate about machine learning, I must balance it during spare hours outside professional work and family commitments. Therefore, please review the answers on this page first. If needed, you can reach me via Messenger at Rakibul Hassan or call 01713095767 (5:00 PM – 6:00 PM).
Thank you all.