---
title: 'AI Engineer Roadmap: How I''d Learn AI in 2026'
source: 'https://youtube.com/watch?v=zwUSZD3t_BU'
video_id: 'zwUSZD3t_BU'
date: 2026-07-31
duration_sec: 3955
---

# AI Engineer Roadmap: How I'd Learn AI in 2026

> Source: [AI Engineer Roadmap: How I'd Learn AI in 2026](https://youtube.com/watch?v=zwUSZD3t_BU)

## Summary

This video presents a practical, job-market-driven roadmap for becoming an AI engineer in 2026. The creator analyzed 700+ real job postings and combined them with hands-on experience from AI projects to design a 33-week study plan with free resources. It covers salary expectations, role categories, required skills, and an actionable week-by-week learning schedule.

### Key Points

- **Roadmap Built on Real Job Data** [00:29] — The creator and his team analyzed 700+ latest AI engineer job postings from different portals to identify in-demand skills, then combined this with their own experience working on 25+ AI projects at ATL.
- **AI Engineer Salaries in India** [01:44] — On Naukri.com, the majority of AI engineer jobs in India pay between 6 and 15 lakh per year. A small number of senior director roles reach 1-5 crore, showing high earning potential.
- **US Salaries and Compensation** [02:55] — The median compensation for AI engineers in the US is $152,000, with big tech companies in Silicon Valley offering $400,000 to $600,000 per year.
- **Three Categories of AI Engineers** [06:38] — AI engineer work falls into three archetypes: Integrator (deploys and integrates models), Builder (builds models, including research engineer and applied scientist), and Allrounder (mixes both, common in small companies).
- **Top Three Skills for Hiring** [11:26] — Karan, AI head and CEO at ATL Technologies, looks for: breadth of AI knowledge including basics and statistical ML, quick prototyping ability, and strong communication skills.
- **Hybrid Problem Solving in Real Projects** [18:07] — For a log classification project, the team used a hybrid solution: 80% via regular expressions, 15% via statistical ML with BERT embeddings, and 5% via LLM classification, balancing cost and explainability.
- **AI Engineer Skill Matrix** [21:00] — The skill matrix maps required skills to role categories: integrators need MLOps/DevOps, builders need deep ML, but everyone needs GenAI skills like RAG, vector databases, and agentic AI.
- **Week 1: AI Basics and Python** [22:52] — Learn the difference between statistical ML, deep ML, NLP, GenAI, and agentic AI, then study Python fundamentals using a free YouTube playlist.
- **Build LinkedIn from Day One** [24:34] — LinkedIn is like a busy street where recruiters notice you. Start building a professional profile immediately with a provided checklist.
- **Week 5: Data Skills and SQL** [39:23] — Learn NumPy, Pandas, and SQL basics. Most of the time your data lives in SQL databases, so knowing basic queries is essential.
- **Statistical ML Is Still Highly Valid** [42:41] — Statistical ML models are lightweight, cheap, and interpretable, making them essential in finance and healthcare due to regulatory requirements.
- **Weeks 12-13: DevOps, MLOps, FastAPI** [46:46] — For integrator roles, learn FastAPI to build a backend around trained models, plus CI/CD, Docker, Kubernetes, and MLOps for deployment and monitoring.
- **Deep Learning and PyTorch** [53:08] — Devote 4 weeks to deep learning fundamentals. PyTorch is the preferred framework, and building an end-to-end project like potato disease classification is recommended.
- **GenAI and Agentic AI** [57:35] — Learn LLMs, vector databases, embeddings, RAG, LangChain, LangGraph, and CrewAI. LangGraph is highlighted as the de facto agentic AI framework.
- **Cloud Platform and Optional Skills** [01:01:33] — Learn at least one cloud platform (Azure or AWS). Optional no-code agentic tools like n8n, Make, and Zapier can help in small company roles.
- **Effective Learning Tips** [01:04:31] — Watch one tutorial, then spend more time digesting, implementing, and sharing. Group learning and consistent action are more effective than jumping between videos.

### Conclusion

The ultimate takeaway is to start with a structured, data-backed learning plan, focus on core fundamentals plus emerging GenAI skills, and simultaneously build online credibility through LinkedIn, GitHub, blogs, and community engagement. With 4 hours of study per day for 6-8 months, anyone can position themselves for a high-paying AI engineer role.

## Transcript

Every month there is a release of new SDK framework. Big tech companies are releasing LLMs every few months. And if you want to become AI engineer, you will have this natural confusion. What skills do you need to learn and in what order?
Now when you go to YouTube, you will find tons of road map videos. Some of these road map videos are created by people who themselves don't have any AI experience. When I started building this road map, I wanted it to be based on
absolute reality. So we did an interesting exercise. We analyzed 700 plus latest AI engineer jobs from different job portals and created the list of hot skills which are in demand. So these are the skills that employers
are looking for as per these latest jobs that we analyzed. We then combined this job analysis with our own experience of working on AI projects in our company ATL. ATL is an AI and data services company which I and my brother founded
and in last 2 years alone we have worked on 25 plus AI projects mostly from small to mediumscale enterprises based in US. As a result we have prepared this practical road map with weekby-eek study plan, free learning resources and
anything unrealistic to increase views of this video. This road map will require 4 hours of study and a time period of anywhere between 6 to 8 shortcut, this is not the right place for you. You can leave this video right
now. Before jumping into the road map, let me discuss few things very quickly. I want to talk about salaries. I want to talk about different category of AI engineers. This information will be very useful to you before you get into the
road map because based on what category of AI engineer you want to become you'll be learning different skills. Okay. On the screen I'm sharing uh the some of the salaries for AI engineers. So if you go to no.com you will find this range
where the salaries are anywhere from 0 to three lakh all the way till 5 K. You see this you you get salary up to 5 cr if you are a senior director etc. And
this is the histogram that I created of real data. This is the real data folks for all the jobs of AI engineer on no.com which is an Indian job portal and you can see that majority of the jobs are in this range 6 to 10 lakh and 10 to
15 lakh. So you can say that 6 to 15 lakh is a common salary range and we found some 89 jobs which had a salary range of 1 to 5 cr and these are senior
directors etc. But if you start becoming AI engineer, this at least shows you the potential that if you have skills, you can reach this kind of financial can reach this kind of financial rewards. And then in US, the median
rewards. And then in US, the median compensation is $152,000. Silicon Valley for some big tech companies, you can get salaries in a companies, you can get salaries in a range of $400,000 to $600,000 a year. I
Valley. So whatever I'm telling you is based on the real information. If you want to know how many jobs are available then once again you can go to any job portal and find out. No is showing 37,000 AI engineer jobs right now. Now
they use semantic search for this. So sometimes you will find some software engineer jobs also in this list but you can rely on this number you know ballpark uh this is a good number. And then for ML engineers there are 111,000
jobs. So this road map is going to be valid for AI engineer, ML engineer, there is a IML engineer. You know companies will post jobs using this different job titles. Even the road map is valid for geni engineer. Okay, genai
is a special category of AI engineer who focuses mainly on generative AI. Now once you decided that okay AI engineer as a career role sounds very good.
Second thing you need to check is whether your natural skills are aligned with AI engineer role or not. And for this purpose we have created this uh suitability test. So what this test will do is it will ask you bunch of questions
on your inclination towards math your learning attitude and so on and based on that it will tell you the percentage matching. It's like a matchmaking between you know boy and girl. This will tell you a percentage matchmaking. It
will tell you whether this career is right for you or not. Now folks, here is the code that we used for analyzing those jobs. And if you look at here for AI engineer jobs, these are the skills in demand. So, Python, machine learning,
AI is a common term. PyTorch, Azure, you can see that Azure is in more demand can see that Azure is in more demand compared to AWS. We did this for different roles. AI researcher for example is a different role. There are
not that many jobs available for AI researchers but it's a very fulfilling algorithm, you like to write CUDA kernel, C++, you care about performance
etc. then this is a right career role for you. Then a IML architect. Then we have uh let's say we have different roles. See data engineer all kind of roles. data scientist. By the way, in this road map, we are not talking too
much about data scientist. Maybe we'll create a different road map for that particular role. All right. So, I'm going to share this code repository with you. And what you can do is let's say you are watching this video 2 years
after this posting you know or after few months. And if you want to do latest analysis, then just use this code and find out the answer yourself. So, here are the tech skills and core skills that you need to learn. As you can see in the
picture, there are many things that you have to learn. But after all, AI engineer career is very fulfilling and they pay you high. So of of course they will demand all these job skills. And then there are optional skills. For
example, cloud AI, reinforcement learning. These are optional skills. Now, if you are a fresher, they will not probably ask you about cloud AI. But person, let's say you are a software engineer, backend engineer, and you want
to learn AI and let's say you learned all these skills. Okay, you also need to have knowledge on at least one cloud either AWS or Azure. Okay, so now let's talk about AI engineer categories. So
although you find tons of jobs with a title AI engineer, when you look at the actual work, it can be divided into three categories. Integrator, builder and allrounder. Allrounder is a mix of integrator and builder. So let me show
you one job post for the first category which is integrator. So here is a job post from Deutsche Bank for AI engineer and if you read the description it says scientists to deploy machine learning models which means data scientists will
actually train the model and then you are involved with deploying those models. Then you will uh design and manage the infrastructure required for hosting ML models. Okay, including crowd resources. you will do CI/CD, docker
things like that. So essentially somebody else is training a model and then you are responsible for deploying it into production integrating it with rest of the systems and of course you need to have knowledge on DevOps, MLOps,
variety of things. The second category is builder category. Now in builder category there are multiple job roles. For example, AI research engineer job role. Now this is the job role for which big companies right let's say open AI
Google all these companies will have this kind of job post they are building this new LLMs they are building novel approaches for training these models so approaches for training these models so they need research engineers even the
big academic institutes like like for example IIT Stanford they will need AI research engineers some of the startups which which are doing some innovative work they will need this AI research engineers. So if you have lot of
interest in math uh performance algorithms C++ etc then you can go for this. The second category of builders is applied scientist. Now applied scientist is a person for example you're working in
Amazon as an applied scientist you will be building the recommendation algorithm for Amazon products. when you go to Amazon website when you're buying any product you see recommended products. So there is an algorithm ML program that
runs behind the scenes and these programs are designed by applied scientist. So for applied scientist they are very much product focused. Okay. So are very much product focused. Okay. So you can say this one is a product
focused ML. Then there is third category data scientist. So data scientist is data scientist. So data scientist is mostly business focused ML. Okay, business focused ML. What I mean by that is data scientists will be hired by even
small companies, banks or healthcare institutes. They want to build custom statistical models. Okay. So data scientists mostly build statistical or
deep ML models. They are good in math, statistics, etc. and they build ML solutions which are business focused whereas applied scientists build ML solutions which are product focused. So these are hired mostly by product
companies. So I'm showing you a job post from Reuters. So Reuters has this senior applied scientist position and [snorts] if you look at the job description see they will usually require PhDs. Okay, you need to have
they will demand a lot of things actually. Okay, you need to be very deep into it. You need to be in a position where you can publish findings on top tier conferences like new IPS. Okay. So you need to have a deep curiosity, deep
knowledge of math, statistics, algorithm etc. And then you will find this kind of job post also. This almost sounds like a data scientist job where you are designing and developing a IML models. Okay, you are doing processing,
cleaning, analyzing data. Okay, then you are developing end toend AI pipelines which means take the data clean it pre-process it feature engineering train the model evaluate it etc and once the model is ready you collaborate with
model is ready you collaborate with software engineers or other integrators to integrate these models into productions okay so this sounds similar to data scientist but they are posting using AI engineer job post so folks all
lot of these jobs will have multiple expectation okay and sometimes employer is also not very clear when they are posting these job post. So when you start working in these companies you will get diverse set of work especially
if you are in a startup the work you will get will contain lot of diversity. Now let me play a clip from my conversation with Karan who is an AI head and CEO at Atl Technologies. What are the top three skills you look for
when hiring an AI engineer? >> That is a good question also. We of course we at Technologies we hire AI engineers. We have been actually hiring one or two right now as well. So if you talk about three top skills that we
would look at while hiring. So the number one would be what is the breadth of the AI knowledge that they have. I'm not talking about them knowing geni. Of course everybody right now is knowing geni for some reason. But how good are
they at the basics? So do they know statistics, probability, math, basic math of course. Do they understand how rulebased systems work? Can they study the data and create their own rulebased systems? Can they like deploy basic
machine learning models like uh like XG boost decision trees or or even simpler models like that? Can they work on their own deep learning model? Can they fine-tune? Can they train? So what I what I look for usually is are they
needed in the industry or they are just like a gen expert. If they're a gen them. If they know things, if they have curiosity to learn more about things, that's a good sign. But only a geni is going to always be a red flag for me.
The second thing that I always look at is prototyping. So right now there is a lot of hype in AI and a lot of use cases fail as well. So prototyping is actually a very useful skill for all the AI engineers out there. So what what's
needed is given a problem, given a use case, you should be able to do like a quick prototype in a couple of days. It could be 1 day, 2, day 1, 5 days. But could be 1 day, 2, day 1, 5 days. But the prototyping would result into uh
like knowing is this a appropriate problem to solve or is this something which is going to like return any results if we solve it or not or how would it like even feel while we are using AI. So it is an immensely useful
skill. We actually train people who like people and expect this out of all our should be good at prototyping when it comes to AI. Otherwise it's a red flag again for us. U the third thing which is countering more towards hype is the
communication skill. You would say that communication is needed in all the roles but in AI the need is even increasing. Why? Because when when you're at asking them certain things or if they're explaining them certain things they
might have already used chat GP for asking those questions or like this creates a divide. Why? Chad Gypy could give you a hallucinated answer or it could give you something which is not taking like all the context in the
consideration. Now if you look at this you need your AI engineer you need your AI team to convince clients to negotiate with them to come to a similar page. Of course the product team would do that but you also
need your engineering team your AI engineering team to come on the same engineering team to come on the same page and do help your product teams. So if you're hiring for an AI engineer or let's say if you're an engineer aspiring
to be an AI engineer you should look at three basic skills the breadth of AI knowledge prototyping and how good are you at communicating things. Whatever Karan said is reflected in the AI engineer job that we have in
technologies. So if you look at this job you will find in the description that we believe that not every problem needs geni hammer. You need to have a knowledge on diverse set of skills. Statistical ML deep learning u see here
Statistical ML deep learning u see here you will see agentic AI system ML and DL frameworks pytor statistical ML ability to prototype quickly and so on. Now this
represents this third category which is allrounder. Allrounder is a person who allrounder. Allrounder is a person who knows statistical ML, deep learning, NLP, geni they know everything. Now you must
be thinking that okay you expect too much. Uh how is that? Well this is true for small companies. Okay, small companies usually need allrounders
and small companies don't expect the person to be very much like highly skilled in in all of this. Okay, it's like a you are jack of all trades. Whereas if you're a builder, let's say if you're working for Amazon and if they
have ML in their job post, they want you to know ML in depth. You should be in a position where you can write even custom algorithm for the ML model. Okay. So the expectation the skill expectation for each of these categories is very high.
Whereas if you work for small companies you need to know this but even if you're average it is going to be good enough and they do that because I will give you short-term project. So let's say there is a 3mon project where you have to use
statistical ML and we have one AI engineer let's say moan is the AI engineer working for ATL. Now for 3 months he's working in statistical ML. After that we may not have statistical ML project. we may have another six
month project which is purely in jai at that time we are not going to hire one more engineer and and moan will sit idle right that that's not going to work therefore in small companies mostly they will look for allrounder whereas in big
companies they can have these two different roles okay so integrator and builder is like a two different roles builder kind of roles you will mainly find in product companies like Amazon, Google etc. Integrated roles are more
common in consulting companies like okay you are working in Accenture okay or cognizant in these companies or let's say big four right PWC etc they will be using readym made model let's say you are using readymade LLM you're using
claude or gemini and you want to integrate it to solve some real business problem okay so here these integrators are good software engineers they know certain ML skills they have good business understanding, they are good in
communication etc. When it comes to allrounder role, you are a jack of all trades. And I compare it with this analogy of a handyman. Let's say you are
a handyman. Now you know that if you want to put a nail in the wall, you need to use this hammer. If you want to remove the bolt from your car tire, you need to use this particular tool. Similarly, a skilled allrounder AI
engineer will have knowledge of all these different tooling including nonML approaches such as rulebased system and they will have this judgment to figure they will have this judgment to figure out what tool to use at what time. Let
me share again a real experience at Atlake. We had a client and we had this project where one aspect of that project was doing a log classification. So this client is a US-based client in finance domain and at first we thought we can
use jai for classifying all the logs and that solution was actually working but it was not the optimal solution because it will incur higher API cost for the
client also there is less AI explanability. So what we did is we identified category of logs where we can use regular expression. So around 80% use regular expression. So around 80% logs had fixed patterns. So if you use
Python regular expression you will be able to classify those. Then for the able to classify those. Then for the remaining categories we found 15% of logs such that they had some patterns. They did not have fixed patterns which
can be captured by regax but they had some patterns which can be captured by statistical machine learning. So we generated BERT encoding and then we used statistical ML approach let's say XG boost or naive base to classify those
and for remaining categories remaining five person where we did not even had enough training samples okay so for those we used LLM classification now as
you can see clearly in this picture we used rulebased system here right rule or traditional programming so this is your tool number one. Then here you use
statistical ML, statistical ML. So this is as a handyman if you're thinking this is your tool number two and then generative AI. So it's a hybrid solution and when you work in any industry project you will not find a
case where you build one fancy model and your entire problem is solved. Usually you have to subdivide the problem into multiple subtask and use different solution for each of these subtask. So
usually you will end up in a hybrid solution. Therefore a good allrounder AI engineer will aim to become a problem solver with strong basics wide tool sets
and the judgment to pick the right tool at the right time. Okay. So here when I say why tool set I I'm not saying okay you learn langraph and then crew AI you just accumulate this bunch of tools you need to have strong fundamentals and you
need to have this judgment on what tool to pick at what time. Now in our road map you will see all these skills and I have built this AI engineer skill matrix where based on the category of the AI engineer that you are aiming you can
focus more or less on these skills. For example, if you are a builder let's say you are building models from scratch. It's okay if you don't know SQL too much. It's okay if you don't know MLOps or DevOps too much because you're mostly
building the new models. Now, if you are an integrator, you have to know MLOps. See, this green means high level of skill. Medium is orange, red is low. But as an integrator, if you don't know
integrator, if you don't know statistical ML or deep ML, it is okay because you will not be building these models from scratch. Now this category jai I'm saying green because you are using
LLM then other than LLM see majority of the companies they don't train LLMs okay LLMs are trained by only few companies in the world it's a whether anthropic or
open AAI or Google etc majority of the companies when they build geni solution they are using rag they are using vector databases therefore I have green okay so here this ji skills is not building LLMs from scratch but knowing the
technologies of geni which is rag vector databases uh agentic AI and so on when databases uh agentic AI and so on when it comes to optional skills see cloud AI doesn't know it's okay but as an integrator and allrounder you need to be
good in it now if you're a fresher it's okay if you don't know cloud AI but if you're good in cloud as a fresher it will give you unfair advantage okay so you can pause this video you can look at the skills. Uh these are the core skills
right? Communication, math, statistics, business understanding and so on. Here is the actual road map PDF. In week one we will learn AI basics and beginners Python. We first need to understand what is the difference between statistical
is the difference between statistical ML, deep ML, NLP, genai, agentic AI all of that and we have this single YouTube video in which this entire AI landscape is covered. So for example statistical ML, deep ML is part of ML. Then you have
few things such as rulebased system etc which comes under AI. Jai, agent AI are the fields which mainly utilizes deep learning. Okay. So this is a 1hour long learning. Okay. So this is a 1hour long video where we have gone over different
concepts right neural network CNN LLM and after going through this video you will have good clarity on the AI landscape then you will start learning
the fundamentals of Python now why Python well didn't you look at the hot skill diagram most of the jobs AI engineer jobs demand Python Python is the top skill when it comes to AI engineer role. So you need to go over
all these concepts and learn basics. We have a YouTube playlist for this. So let me just show you that particular playlist and you need to go over uh certain videos here like the first 16. You can learn Python from other
resources as well. And then by learning Python fundamentals eventually in the code. So this is the code snippet of training a statistical machine learning model using Python. So this is all Python code. So you probably now
understand that whether you are in genai, deep ML, statistical ML, you will be using Python everywhere. Python is like God. It's everywhere. Now during the same time period, you will start building your
LinkedIn profile. We are not going to wait till the end. LinkedIn is extremely crucial for your online credibility and we have provided a checklist. So if you click this particular link here, you will find a checklist. So just go
through this checklist, act on all these items and in the end you will have a professionallook profile. Now why is LinkedIn important? Let me give you analogy. Let's say you know how to make samosa really well and you have this
shop on the right hand side in a village. Now in a village there is hardly any footfall. So you are not getting any business. But somebody advises you to move your shop in a busy street in Delhi and you do that and all
of a sudden your business grows 10x. Why? Because you are on a street where many people are passing. So LinkedIn is that street where there are many recruiters, many managers who who who are present and they will notice your
profile. So LinkedIn is like a busy street where you will get an opportunity to showcase your skills to a wide range of people. And folks, building online credibility, building relationship takes time. Therefore, you should not wait all
okay, let me learn technical skills first and I will do LinkedIn later on. No, you should do it from day one. In week two, you will learn data structures and algorithm. And in data structures, uh it's okay if you learn these many and
you can skip the rest. In the algorithm, I think if you learn one search and one short algorithm, it is good. uh as of this stage and later on as in when you need to learn new algorithms you can learn it in terms of learning resources
I have this playlist here so let me show you that playlist and it has got lot of views see millions of views and just look at all the topics which I have mentioned in the road map for example you don't need to go over merge short
insertion short share sort etc right I have mentioned the videos that you can skip here then in terms of core skills you need to start working on your communication. For this, the excellent resource is Toastmasters. So, this is a
free resource. They have clubs everywhere. So, click on find visit a club and enter your uh city. Let's say you are in Hyderabad. And in the Hyderabad, see there are so many clubs which are present. I have attended these
sessions. Here you can practice your verbal skills. Okay? you will practice your negotiation skills. It is totally free folks. So use this platform to improve your communication skills. And in this same week, I will advise you to
watch this conversation which I had with a senior director of fractal. Okay. And here in this conversation, we even discuss what are the skills that Rishi looks at when he's hiring people. See, he emphasized on three qualities.
Humble, hungry, and smart characters. So looks like these companies they often looks like these companies they often focus more on soft skills than the hard technical skills. In the age of AI coding is done by AI. So the importance
of soft skills have gone up. In terms of assignments, you will find uh the assignments in this same playlist. Okay. So whatever playlist I have here uh I
will have a corresponding GitHub repository [snorts] and that GitHub repository you will find in the video description. So let's see if I go to video description here. See I will find this and there will be an exercise.
Okay. So I have an exercise for Python pretty much everything. See these are the exercises for Python. Let's say read and write file. Okay. So there are nice exercises which are given here. You can also use chat GPT by the way to practice
your concept. Let's say you are learning Python dictionaries. Okay. You can say I'm learning Python dictionaries. Give me one simple coding exercise. And Jet GPT is like this one-on-one tutor that you can work with. Okay. So you can
write a code then you can give that code to Jet GPT. You can ask it to evaluate it and that way you have a conversation and you can work on variety of assignments. In week three after you have cleared the fundamentals of data
structures and algorithm you will learn some advanced concepts in Python. Okay. So you will refer to the same playlist and in that go to video number 17 to 27
where you will learn all these concepts and then there are exercises in the same and then there are exercises in the same playlist. In terms of core skills see as a first step you created a LinkedIn profile professional looking LinkedIn
profile as a second step you will start following some of the influencers. Yan Leon for example he's the person who invented convolutional neural network okay and he writes many good post so what you can do is you can read through
his post and you can also start engaging you can start commenting etc and you will realize the benefit of this commenting process later on but let me tell you it has helped me tremendously in my career and it will help you too
you will find many prominent AI influencers on Twitter actually so there are more active on Twitter than LinkedIn. For example, Andre Karpathi. Okay. So, Andre Karpathi writes a lot of useful post. So, you can read through
it. You can comment on it. Sometimes these guys will even respond to your comments. Remember that online presence is a new form of réumé. Then to improve your business fundamentals, see many companies if you look at job
description, you know, I will advise you to look at all these job post just spend one or two hours just reading through various descriptions that will give you the reality of the job market. Okay? What uh employers are looking for. One
common mistake that people make is they learn some skills by following some random YouTube video and then in the end they go to these job portals. you should understand what employers are looking for and then you prepare for those
skills which are relevant. Okay. So business fundamentals I have seen that I was talking with a director of data engineering for a big pharma company here in New York and what he told me is that they look for people who have
domain knowledge in pharma. So domain knowledge is very important and you can understanding by going over some business case study YouTube channels such as think school they have all these case studies and in that case studies
you will understand accounting principles, business principles and so on. Then you should also start participating in prominent discord servers. For example, for code basics, we have this particular discord server
where you will find there are like 50,000 people, 51,000 people here. And let's say if you have SQL question, you ask here, you have a question on Python, you ask here. Okay? So by interacting with people, you are developing your
communication skills, which is going to be very crucial because when you work in any company, you'll be working in a team and you are interacting with people. So you need to have good communication even you're interacting with business
stakeholder you know sometimes there is this uh discipline to communication how you answer how you become polite how you have empathy towards other people and so have empathy towards other people and so on by participating in this discourse
server you will build another great skill which is the teaching skills. I many times see people that they themselves are learning concepts but somebody else when they ask a question they are ready to explain that concept.
Now this explanation skills are going to be super important. I have looked at some of the job post and they specifically ask for explanation skills company if you are dealing with let's
say non- tech uh business stakeholder now you want to explain certain concept to them. If you're good at explaining things you will be able to convince them and that that skill matters a lot and this is something you have seen in the
video clip of Karan. Okay, at at when we hire people, one of the skills that we look at is is the person good at explaining things? Is the person good at communication overall? When you post question in discord, there is some
discipline, you know, some mannerism that you need to follow. For example, you are facing an error in Python code. Don't just copy paste the error in discord and say that okay, can you help me answer that that question? You have
me answer that that question? You have to be polite and you have to uh showcase yourself as somebody who is looking for troubleshooting not the spoon fitting. Okay. So you can say that I'm stuck here despite XYZ and can you provide me tips
that oh I get this error can someone help. Okay. Now you can also use chat GPT if you're facing error go to chat GPT and say explain me this error step GPT and say explain me this error step by step and try to understand what it is
saying. Okay, in terms of assignment, you will write 10 meaningful comments in AI related post and then note down your key learnings from the case studies at think school or some other uh YouTube channel. In terms of motivation, I have
this video that you definitely need to watch. So this person is an uh ML watch. So this person is an uh ML engineer at April right now. He has a mechanical engineering background and when he was in college he used to
participate in kaggel. So using kegel contribution he built this online credibility and he directly got ML job without any formal education in ML. Okay. So this interview was recorded 4 years back. I met him uh during our AI
fest when I went to India last time and I learned that he's now working at Apple as an ML engineer. So just imagine how important is your kegel contribution. Okay. So you can go to kegel just in case if you don't know what kegel is. It
case if you don't know what kegel is. It is a platform where you can practice different [snorts] machine learning competitions. Okay. So if you go to competitions there will be AI competitions. For example, let's say
house prize. Okay. Llm classification finetuning and then you will come to finetuning and then you will come to this leaderboard etc. Okay. So it's like you're playing a game. Okay. And they also have data sets. Okay. They have
variety of useful resources. So please explore that. So that is the end of week three. In week four you're going to work on version control. See when you collaborate on code you have to use this platform GitHub. And underlying GitHub
is this version control system called Git. you need to know basics of uh version control system, basic commands, uh pull requests, etc. And I have provided some playlist here so you can refer to those. In terms of core skills,
I think presentation skill is probably one of the most important core skill and I have this video death by PowerPoint and it's like a bible of presentation. So if you see the video and if you just follow the guidelines in that video,
you'll be amazed at the result. So say you are in college and if you are presenting okay if you have worked on project and if you are presenting use the principles which are stated in this particular video okay so I think he gave
a TED talk uh I think this is a TED talk yes excellent folks I mean he has said some simple techniques but I see majority of the people they don't follow it when they are creating presentation in terms of assignment you will write
in terms of assignment you will write two meaningful blog post on AI topic. Okay, for example, how CNN's work. Now, there are many different platforms where you can write blog post. One of the guy I know that he got a huge success by
writing blog post is Himmanu Dubet. Okay. So, he was a student and he had this habit of uh being active on Twitter. So, let me show you his Twitter
profile as well. Once again folks, many of the tech people, good tech people in AI industry, you will find on Twitter. Twitter is like the de facto platform. So it will help you if you are interacting with those people via
Twitter. So Himmanshuh is writing this blog post. Okay. So he has his own website and if you look at this post, he has post on a IML etc. Right? Like initial thoughts on llama. So whatever article you are reading, you can even
explain okay how CNN works. I think he has explained some of those things right CNN from scratch with pure mathematical intuition. So now when you write this blog post and if somebody notices your work they will approach you. In case of
Himanchu he was approached by some startup in the US and I think recently he visited San Francisco also. So he's not from like IIT or some big college.
Okay. He's from I think he's probably from a small town and from a small town and he uh just focused on building his he uh just focused on building his online credibility and how did he do it?
He started writing blogs. Blog writing is easy. You can use even AI tools for your English correction. Definitely add your thoughts, okay? Don't just copy paste. And then Twitter. Twitter is extremely important. Okay. So that will
be your assignment writing two meaningful post and see if you look at any job post folks see look at this job post effective communication with key stakeholders stakeholder management communication is the fundamental skill
communication is the fundamental skill that every single AI engineer job will be looking for. Therefore you can't uh you know underestimate it and I have given profile of some other person as well. So I'm seeing a lot of these
people who are let's say from small town in India they are from tier three college but they have figured out this art of online credibility okay and they participate and for example this this person he's a research collaborator at
coad which is a very good company right and look at his his education background you can check his background and all these people come from very humble background okay tier three college small town and they get job immediately after
their engineering. They don't have to even apply because somebody has noticed them. In week five, you will cover numpy, pandas and data visualization. Numpai and pandas are the Python packages that you use to perform
exploratory data analysis to perform data cleaning. These steps are required before you start training your model. You will also learn SQL basics. Now see
you don't need to go in depth for SQL as long as you know basics like the basics queries you know select where disting etc the basic joins you know the primary
key foreign key some simple basic concept should be good enough because as an AI engineer sometimes you will be interacting with SQL databases I I I think most of the time your data will be living in some SQL database so when you
want to pull it to train your model or to build your rack solution whatever you will be writing those SQL queries and that is the reason SQL is added here. So that is the reason SQL is added here. So for SQL I have this YouTube tutorial
very simple 1 hour video is good enough and for numpy and pandas in my course I have made those chapters free. So see numpy chapter all the videos are free
and then pandas mattplot lib all the videos are free. So you just watch it and that should be good enough. Pandas is a vast topic. You don't need to learn every single thing in pandas. Okay, just learn these basics and start doing your
work and in the future if you need to learn extra concepts you can learn at that time. Week 6 7 8 you will focus on math and statistics for AI. If you look at any AI engineer job you will not find math as a skill but if you go for any
math question. Okay. So it is kind of assumed that you have some fundamentals on math and statistics. Now if you are on math and statistics. Now if you are targeting integrator role then it's okay
not to go too much in depth in math okay some basic math is good enough but if you're targeting a builder role or let's say allrounder role then all these know now you'll be like okay but nowadays we use LLM why do we need to
use this well I talked about that at real project experience right where for statistical ical ML and when you use statistical ML or deep ML in order to
evaluate the model in order to pick a right model you need all these fundamentals. So folks trust me I work on industrial AI projects all these fundamentals are going to be useful either today or tomorrow and when you go
to interview they will definitely ask you these questions and by the way some people have this uh thought that okay they're not good in math but if you follow right resources for example Khan Academy three blue one
brown okay I hope you have seen this channel three blue one brown state Quest. These are amazing channels and even if you got let's say 10 out of 100 during your school days in math, if you watch these amazing teachers in math,
you will be like okay I can do math. You will be filled with that confidence. Okay. All you need is a good teacher. So please try it out. Math is seriously not that hard. Okay. And then there are exercises that you can follow. Week 9,
10, 11 will be statistical ML. Now somebody asked me this question that hey somebody asked me this question that hey the we live in this LLM era so LLMs can do classification they can do regression so how relevant is statistical ML I gave
so how relevant is statistical ML I gave that answer by giving the analogy of a Bangalore traffic let's say you are going in a Bangalore in a crazy traffic will you take a bike or will you take a big car most of the time people will
take bike why because it is lightweight weight. It will help you navigate faster in the traffic. It will save you fuel cost. Similarly, statistical ML models
are lightweight. They are very cheap. In some domains such as healthcare and finance, in all these industries, there is a high requirement for And with statistical ML, let's say if you're using linear regression or some
simple model, you will get high interpretability. I have a friend who works as a data scientist in some fintech company and when she builds a model they prefer statistical model over other complex models you know like genai
or deep learning because in statistical model which are simple you get good explanability and due to regulatory requirements you need to have that so requirements you need to have that so statistical ML is 100% valid nowadays
statistical ML is 100% valid nowadays folks we have clients in ATL we use statistical ML Now in terms of learning there are two big modules pre-processing and model building. Pre-processing means making
your data ready for model training. And this includes handling NA values outlier treatment data normalization encoding your data feature engineering your train test split and so on. When it comes to model building you'll be mostly
working on regression or classification. Okay. And for linear models you can learn linear regression uh gradient descent logistic regression. For nonlinear models you can learn decisionry, random forest and xg boost.
Uh in our practical experience we mostly use xg boost and random forest. They are very efficient. They give you high accuracy. In terms of model evaluation, you need to know all these parameters. Okay. How to evaluate the model. Now
model evolution is very important folks because that will help you figure out if model is good to go in production. Hypertuning topics these are the listed topics for hypertuning and when it comes to unsupervised learning K means DB scan
etc is enough. Now we have this playlist in YouTube it has got see millions of views. Uh so you can just follow this playlist. I have highlighted how many videos you need to watch. And then in
terms of core and soft skills. See when you get hired in any company, you'll be manager. They'll be doing project management through uh these techniques. training series. So when I was at Bloomberg, we had an agile coach who
came and he forwarded this particular material. These are bunch of free videos, excellent videos which will tell you what is scrum. So scrum is a common technique that they use for project
management. And in terms of tools, they use Jira and notion. Okay. And these are the exercises folks. So I have given exercises in my ML playlist. You also exercises in my ML playlist. You also need to work on two kegel ML notebooks
and then write two LinkedIn post. Whatever you have learned in ML, just write a LinkedIn post or whatever projects you have built. Let's say you worked on some kegel notebook. You can share your learning via LinkedIn post.
Okay. In discord at least help 10 people with their questions. See helping people will consolidate your own knowledge. It will give you an opportunity to practice your explanation skills. It will also help you build relationships. Okay. So
there are numerous benefits that you can get out of it. Week 12 and 13 you will learn DevOps, MLOps and fast API. See if you get an AI engineer job which is falling under integrator category. Then you will take a trained model and you
will deploy it. For deployment you need fast API. Okay. So fast API is a popular Python framework where you can take a model and you can write a backend server. See model is like our human brain. But just by having brain you
can't get things done. You need a body. So right now I'm talking I have brain but I have eyes, I have mouth, I have hands, I have body. So fast API will help you build that body around train model. And then to deploy the model you
need to know CI/CD pipeline, docker, kubernetes, all of that. MLOps is required for experiment tracking, monitoring your model performance in production etc. And then you need to have familiarity with at least one cloud
platform. Now there are three popular platforms AWS, Azure and GCP. Out of platforms AWS, Azure and GCP. Out of these top two are AWS and Azure. And it is clearly reflected in this analysis job analysis. See Azure is at the top
and then there is AWS. I have provided a link of free YouTube tutorials for each of these topics. So please refer to that. Now in this time period you should
start contributing to opensource. You'll have this question that okay I'm still learning. I'm not an expert. How can I contribute to opensource? That is totally a wrong concept folks. You can contribute to opensource. You can go to
some repository, fix the documentation or fix an easy issue. So let's say if you go to GitHub. So let's say GitHub hugging face repository. Okay. So hugging face is a open-source package. You can go to their issues. So they will
have let's say thousand issues pending which means there are thousand pending bugs that they want to fix. And if you go to label and search for uh good first
issue. So these are the beginner friendly issues I have labeled here. So good first issues you'll find tons of issues. If you want to target simple repositories then see mine SQL is one repository that you can target. There is
another repository called rag rank. So just find all these repositories and start contributing. Okay. So you go to issues look at this issue and then folks use your friend. Who is your friend? Well chat GPT. So you can say that go to
Well chat GPT. So you can say that go to chat gpt and say I want to contribute to opensource guide me on first steps. See it will guide you step by step folks and if you are looking at any issue you can say okay
me fix it? It will fix it. It will also tell you what to do what what steps you need to follow. Chat GPT is likeading you know it is there in your service all the time. You just need to know how to use it. All right. Now, the reason I am
putting so much focus on opensource contribution is look at this particular video. Okay. This person got two ML engineer jobs immediately after 12th. He did not even go to college. How was that possible? Well, the repository which I
showed you, rag rank is written by this guy. Okay? He's from small town in Kerala. It's not like he's from IIT and from some big town and big connection. from some big town and big connection. No very simple background but he
understood that open-source contribution is a way to go. So watch that video you will get lot of good tips. All right in week 14 you will build machine learning projects and I will say okay build just one project which is uh this particular
project where you are going over all the steps in uh ML model development. So you are doing data cleaning, feature engineering, outright removal, see model building, you're also writing Python flask server. So flask is another
framework similar to fast API. I would advise you to use fast API uh instead of flask because I think it's better and then you are building website, you are even deploying it to AWS. Okay. So all these steps you will cover and you will
these steps you will cover and you will build an end toend project. During this same time you should start building your resume. Okay. And I have provided some good videos. I have also provided you this checklist. So if you click on this
checklist. Okay. See this is a checklist. So if you follow all this checklist you will end up in an excellent ATS you know application tracking system ATS compliant resumeum. You also need to start building your
project portfolio website because see you build this project then you create a website. So let me just show you this particular website. So this is the website that all our boot camp students get. You can create portfolio by going
to GitHub. There are some free tools available as well. So here this person has given idea on what projects he has worked on and see you can see his projects here along with the screenshot. You can go to his GitHub, you can check
You can go to his GitHub, you can check his code, you can um I think look at his LinkedIn profile where he posted about his project. Okay, so this is a project I did uh and so on. So this project portfolio website is sort of like your
live resume. A recruiter will be really impressed if you have a professional looking profile. In terms of assignments, as I told you already in that project instead of flask use fast API. You can also build one more
classification project. So this is a regression project but I will advise you to build a classification project. You can get data sets from Kel from variety of websites. Okay. And then in your project portfolio website now you have
regression project which is going to be super amazing. All right. Next step is deep learning. So you'll spend four weeks in deep learning. See deep learning is the reason behind modernday LLM boom. So deep learning is essential.
you questions on neural networks, different architectures, forward propagation, back propagation and so on. Now there are two frameworks. TensorFlow, PyTorch. PyTorch is number one folks. You should definitely learn
PyTorch. TensorFlow is something it's little bit low level and in some jobs they will ask for TensorFlow. Now I have this particular playlist [clears throat] where I have taught things using TensorFlow. I'm going to build a new
playlist using PyTorch. But this playlist is not just TensorFlow. It is videos here which will go over fundamentals right like what is neuron, how neural network works, then u activation function, derivatives. So I
have gone into maths in detail. So look at this playlist for fundamentals. If you're interested in learning PyTorch then you can refer to this campus playlist. I think he uses English and Hindi. So if you know these languages
this is an excellent playlist that you can follow. And then for end to end can follow. And then for end to end project I have this amazing project uh where you are building a disease classification for potato plant. Okay.
So if you look at this project what we did is we created a mobile app. So look at this mobile app. Okay. In this mobile app you go and you take a picture of a
plant and then it will send that picture to your fast API back end and back end will have a trained model. It will do the prediction. It will send the result back to your mobile app which is returned in react native. So this is see
early blight 100%. So this is an amazing end to-end project that you can add to your rumé and of course you don't want to just copy paste this project right you can use a different data set instead of potato plant use some other plant
images and deploy to Azure instead of GCP okay so you can make customization GCP okay so you can make customization and make it your own unique project after you are done with the project you will create a presentation and you will
present it to stakeholders and put it on LinkedIn Okay, week 19 2021 you will LinkedIn Okay, week 19 2021 you will spend in either NLP or computer vision. What I have observed is there are some AI engineer who will focus on text which
is NLP. There are other engineers who will focus on images which is computer vision. Okay. So you can choose one path. Uh I would say you don't need to go both both the ways. Okay. Start with one. Uh in terms of NLP you will learn
regular expression then text representation. These are the topics and once again I have a playlist that you can refer to. This playlist has got very can refer to. This playlist has got very good uh response okay and I have covered
good uh response okay and I have covered all the topics in very much in detail. all the topics in very much in detail. Now for the project see in this playlist you we have a project using dialog flow but dialogflow is kind of getting older
and there are new technologies coming up. So I would say skip this particular project instead of this do this project. Okay. Now this is the same project which I explained at the beginning of the video where for
you you remember this diagram I hope you remember it. So we actually implemented this particular project where we used a hybrid approach of uh regular expression hybrid approach of uh regular expression statistical ML and genai. Okay. So do
this project and then for computer vision uh in ATL once we had a project vision uh in ATL once we had a project where some grocery shops in US they wanted us to build a solution where you take a picture of the items which are
kept at the shelf and it should count those items. For example, Cheerios. Okay, this item is Cheerios. So it will do object detection and then it will do the count as well. Okay, so this is the use case. Uh these are the things you
will learn. OpenCV is the library that you will learn. Then you will use some of the concepts in deep learning. For example, convolutional neural network. It is heavily used for image processing. So you can use that. You can also learn
YOLO and all these other techniques. Okay. And at the end comes the most interesting part or the part that everybody is waiting for which is JAI and agentic AI. There is no doubt we are living through a boom and jai and
aentici is a very very hot skill nowadays. In terms of topics you need to know what is llm vector databases embeddings. What is retrieval augmented generation? Lang chain is a de facto framework that people use. People use
framework that people use. People use langraph for agentic uh AI. Crew AI is another agentic AI framework. I have provided all the resources. See for langraph for example I have this complete crash course okay complete
crash course on lang graph crash course on QI folks this information is there on YouTube for free if you have motivation if you have discipline you can learn it for free without spending any money then MCP then we have complete ji crash
MCP then we have complete ji crash course this is a I think this is 3hour long uh YouTube crash course which is available uh for you for free and then week 25 to
27 you will spend in building the projects. See you learned all the skills but until you build end to end projects your skill building process will not be complete. So you need to target some projects where you are using rag you're
using some agentic AI etc to solve real life problems. So I have this playlist and I'm adding new projects actively. project you already did, right? But this this these two and then aentic using
langraph. Okay. Okay. So there are I think this one two and three this project you already did. So these projects are good enough and make sure you add these projects to your project portfolio. All right. Now in week 28 and
9 you need to work on unguided projects. So for that uh a very good resource is uh this data challenge. So in at code basics we conduct these data challenges.
These are free there is no money. Here we post a problem statement along with data set you know sort of like kegle and you build a solution. After you build a solution, you will make a presentation and you will present it uh via LinkedIn
post. So this is how kegel and these challenges differ. Kaggel is just building technical skills but here you are building both technical skills as well as soft skills. So Arin Sharma is a person who won one of our data
challenges and he got a job based on that. Okay. So let me show you. So this that. Okay. So let me show you. So this is the challenge that he won and when he presented see if you look at his presentation he's presenting as if as if
he's talking to business stakeholders. Okay. And this was noticed by one of the recruiter and he got the job opportunity. It's like he did not apply reached out to him. So this
participating in these challenges and winning the challenges can be very beneficial. So the first challenge is building a rag based assistant which will be controlled based on the roles. Okay. So there is role based access
control RBAC in this project. So it's not a simple rag project. It is bit complicated. It is very much similar to how you will build the project in the industry. The other challenge for which we announced
other challenge for which we announced the winners just today by the way is the winners just today by the way is this uh challenge to use NLP to detect adverse drug effects. Okay. So this challenge is actually very complex and
if you can do it it will be amazing. So we conduct these challenges we announce the winners but even after the challenge is closed you can still practice it. The problem statement is there, data set is there. So practice it and then add this
project to your project portfolio website. Okay. Now week 30 to 32 you will learn at least one cloud platform. I will say go for either Azure or AWS.
I will say go for either Azure or AWS. Now as a fresher you are not required to learn this. So if you go for any fresher interview let's say person comes to a fresher interview in our company at lake. We don't ask any cloud question.
But if a person knows cloud then it will give them unfair advantage. So based on the time and willingness that you have you can learn. If you are an experienced professional you need to know it. I'm seeing this trend nowadays where many
software engineers want to learn AI and want to move into this. Okay for them this integrator role is a perfect. So you already know back end, you already know one of these cloud platforms. You need to know certain ML concepts and now
you can become AI engineer which falls into that integrator category. I have mentioned the topics for both Azure and AWS. And folks see once you learn one topic for example identity and access management AM right if you learn it in
AWS learning it in Azure is going to be super easy. So you will learn this transferable skills. So just learn one cloud platform and when you work on other cloud platform you know the fundamentals. So you'll be able to learn
the Second Cloud platform very easily. Now I don't have any specific recommendation for this resource. You'll find tons of YouTube videos, tons of free courses. So go figure it out at some point on our channel. We plan to
publish videos on these cloud platforms as well. Now the optional skill is no code agentic tool. NAN is very popular. Make Zapier these are no code agentic tools. Okay, this is an optional skills.
There are most of the AI engineers jobs will not ask for it. But let's say if you're working for a small company where the clients projects are demanding both of these you know no code tools as well as uh coding frameworks like langraph
then in their job posting you will find it. So after you have learned all the uh mandatory skills if you have time you can learn at least one of these two can learn at least one of these two tools and then week 33 and onwards more
projects online building through LinkedIn kegel discord opensource contribution and then eventually someone will notice you folks and you will get will notice you folks and you will get the job success. Now I'm going to build
the job success. Now I'm going to build another video for uh how to apply for jobs and how to get success because there is a science to it. I will talk about that funnel where you apply then you get shortlisted, you appear in the
through the entire process. So that is going to be the follow-up video. Uh so please stay tuned uh watch it out. Tips for effective learning is going to be don't watch 10 tutorials on the same topic. Watch one tutorial, spend time in
digesting, implementing and sharing. You know, spend 15 minutes watching the tutorial, 45 minutes in digesting, implementing and sharing. If you're jumping from one video to another, you will drain out your energy. It's not
Okay? So, it's a simple principles requires lot of mental discipline. You almost need to put your phone away when you're learning and you need to have this discipline that I will spend more time in these three and less time in
consuming. Group learning is also very effective. If you're learning let's say effective. If you're learning let's say uh swimming and if you go alone versus if you go with two of your friends. I think doing it in group is going to
motivate you more. It is going to give you more inspiration. Similarly in code basics discord server you will see this partner and group finder channel where
people will be like hey can anyone join me in learning AI agents and you can respond and you can kind of make a study group watching the video is not going to be enough you need to act on it folks I have provided all the resources in the
have provided all the resources in the video description please take them start your action from today I wish you all the best and if you have Any question the best and if you have Any question there is a comment box below.