Day in the Life of an AI Engineer

Career Exploration
27 followers
4 Video Views·Aug 25, 2026
In this video, we break down the daily routine of an AI Engineer in a simple and practical way — from working with data and building models to collaborating with teams and deploying AI solutions.
Whether you're planning to become an AI Engineer or just curious about the role, this video will give you a clear idea of what actually happens behind the scenes.
Do you want to learn technology from me? Check https://resources.codebasics.io/kbKar3 for my affordable video courses.
Need help building software or data analytics/AI solutions? My company https://www.atliq.com/ can help. Click on the Contact button on that website.
🎥 Codebasics Hindi channel: / @codebasicshindi
AI Generated·May contain errors
I am a developer and I'm very good at Python and suddenly after 4.6 release, my skills
have no value.
Everyone sees AI as a magic, but it is just a tool.
So I need to now start, how can I unlearn a few things?
How can I remove the baggage I'm carrying as a developer?
Welcome to Atleek Technologies.
Atleek is an AI services company and we have a team of AI engineers here.
Today, I'm going to show you a day in the life of an AI engineer.
I'm the co-founder of Atleek and my involvement here is in sales, dealing with clients and
participating in the technical architecture for AI projects.
I also do video shooting for my code basics work right from this office.
The AI engineering team sits in the next office.
So we are going to walk over there.
Let's go.
Hey Shikant.
Yeah.
Hi.
How are you?
I'm good.
Great.
Shikant, you want to introduce yourself?
Yeah, sure.
My name is Shikant Dongala and I'm working as an AI engineer at Atleek Technologies.
Okay.
So we would like to know what kind of project you are working on and what tech stack do
you use?
So currently there are a lot of projects that I'm currently working on, but one project
to pick is in the fintech domain.
So which is about analyzing the spending patterns of customers and then helping them to improve
their credit score, create insights, and then building recommendation engine so that they
see the actual products that are highly relevant to their spending nature and spending patterns.
So this is in the fintech domain.
Okay.
And what kind of technology and tools do you use here?
Most of the things backend is in Python and for the cloud we are using AWS and we'll have
a load balancer and Lambda functions and we are using Fargate to manage all the load and
concurrency handlers.
So this is mostly, and we also use Docker, we'll use Postman, you know.
And in terms of LLM, like which LLM do you use here?
So we use OpenAI and also we are now recently testing the server models to test the capabilities
of Indian models and see how it goes.
Oh, wow.
That's pretty amazing.
Okay.
I want to know how does your day look like, typical day look like from morning till evening?
These days, my day starts around 7 or 8 a.m. and I spend some time in researching, collecting
most important resources that I want to, you know, do research on or to look into for that
particular day.
Because every day new things are being released, right?
Yeah, it is very difficult to, you know, catch up with everything, but you'll have to try
to do as much as you can.
Yesterday I started setting up OpenCLAW and when I was on second step, I heard a news
It's crazy.
So it's difficult to, you know, catch up with everything, but we'll have to try as much
as we can.
So I will start my day with, you know, identifying what all things that I need to cover for that
day.
And I will take short break for one hour, one and a half hour.
And then I will be back in office by around 9.30 or 10ish.
And then the very first thing I do is check my emails and messages.
And I will complete that routine for about 5 or 10 minutes.
And then I will start planning my day, what all tasks that I need to do for that particular
day.
And once I do that planning, I will pick the most complex task for that particular day
because in the morning we'll be fresh.
Oh, that's a good idea.
So I will, you know, start with the most challenging task for that particular day and I will be
on it for my entire first half.
And in parallel, we'll also having some internal, you know, team sync ups and scrum calls.
So that will be the priority for the first half.
And I will try to keep most of the calls in the second half of the day because, you know,
to make most of, to, you know, better manage our time.
So in the second half, we'll be having some scrum calls, internal sync ups.
And again, these are the days where we alone can't keep up with the research work that's
happening, right?
So we'll have some internal sessions where we'll be, you know, sharing our knowledge.
So everyone will be updated on that along with some client calls and everything.
Okay.
And I'm seeing some, something interesting, like, is this like a technical architecture
diagram?
Yeah.
So this is a basic flow of this particular, you know, data starting from the ingestion
and how we process all the transactional information and how it helps, you know, generate insights
for the users, understanding their financial nature, financial patterns, and then help
us recommending some good products based on the user spending patterns.
And in terms of framework, do you use a lang chain or something else?
I use lang chain.
And again, it depends on the use case.
Sometimes I also use crew AI when I need to, you know, have multiple, you know, agents
working on, then have a better agent to orchestrate that.
But as part of the, you know, framework, I'm mostly using lang chain.
Okay.
And we'll also have lang smith, lang graph, depending on the things.
So how often do you interact with clients?
Do you have a separate product management team?
Yes, we do have separate product management team and here they are.
Hello.
So they are the driving force.
Okay.
And you have any like timing related situation?
Because most of the clients are based in US and UAE.
So you have to take calls in the evening or something?
Yeah.
So we manage our workload.
So whenever we want to, you know, stretch our hours, we have to attend, you know, calls
at the late night.
Right.
So we will manage our day accordingly.
And then you get some break in the afternoon as well.
Exactly.
And this is our activity area.
Okay.
Where we will be, you know, having all the presentations going on, key discussions happening
in this area.
Okay.
So I think we covered till afternoon.
So how does your day look like after lunch?
So after working on the complex task, right?
So I need some refreshment.
So we have, we are flexible.
So in the afternoon, I go my home, which is nearby, I'll have my lunch and I will take
a break for about an hour.
And then I will be back in office again.
And I will again start organizing my day for the second half.
I will start with the important tasks again, do some coding and complete a few tasks.
And then we'll also have some internal meetings.
So I will try to keep my internal meetings at the very end of the day.
And so after lunch, we'll always, you know, be a bit sleepy, right?
So we have the flexibility to take a nap as well.
And we have some coffee breaks.
We go out to get some refreshment.
And these are essential to keep yourself, you know, refreshed and be active when you're
working on something very complex, right?
So we have the flexibility and we'll be back to our client meetings or internal meetings
at the end of the day.
And my day ends at around 5.30, 6 o'clock.
Okay.
Good.
Now, so I would like to ask some general question because you have been working as an AI engineer
for how many years now?
Last three years.
Three years?
Yeah.
Okay.
So you have seen days of statistical ML all the way till GPTs and Gen AI, all of that.
Many of the viewers who are watching this video will be beginner or they have aspiration
to become AI engineer.
So can you tell me what is that one mistake that every beginner makes, which you might
have made as well and you learned along the way?
Yes.
So one thing that any beginner should remember is do not jump to a framework or a technique
or a model directly.
First understand the problem, brainstorm it, and then pick the right solution for that
problem.
Because a few problems you can do it with statistics, with ML, and with Gen AI as well.
So understand the problem and then pick the right tool that you actually need.
Okay.
Instead of jumping directly with Gen AI or ML.
That's what I usually say, like don't marry to a specific tool or technology.
When people say I'm a .NET developer, I'm a Java developer, it really freaks me out,
you know.
We live in the vibe coding era.
You can code in anything.
So by the way, since we mentioned vibe coding, how much of vibe coding do you use in your
day to day work?
That's actually a very good point to bring.
And so recently, it's like 40 to 40 or 50% we can say for the AI specific tasks, whenever
we want to create some boilerplate, some pipelines, right?
So mostly we use AI to do that automation.
And then we use that time to work on some complex tasks.
So be it architecture, be it in terms of security, safety.
So we utilize this time, that day.
So these are the days where we don't spend just on working on AI.
So we take additional tasks for software engineers as well, right?
So for that, I think it can do like 70-80% of the code is vibe coding.
But for the AI specific, it will be 30-40%.
And yesterday I noticed one example where we were parsing the document using DocLink
And DocLink's default parser has some issue parsing the headers, you know, it doesn't
provide the hierarchy correctly.
Now if you're writing code using cloud code, it will just use whatever is the default.
But then we went to GitHub, we found that it is a known issue.
And somebody has provided some solution, some workaround, we ended up using that workaround.
And that's where we felt that as an AI engineer, our jobs are safe because cloud code cannot
fully automate it.
And it was an interesting perspective that we built yesterday while working on this real
problem.
That's great.
Okay, so what is the fun part of your day?
So it is convincing people what AI cannot do.
Because they think it's magic, right?
Yeah, so everyone sees AI as magic, but it is just a tool, it cannot do everything that
you wish for.
It has some limitations, it has its strengths.
So we'll have to leverage those strengths and we'll have to back AI up with our human
intelligence.
So I think convincing people what AI cannot do, that's the most fun.
And how often do you interact with clients?
I know that there are product managers, but you also interact with clients directly, correct?
We do, because there will be a few technical clients where we will be dealing with them
directly and a few projects will be handled by the product managing team.
But we do interact with clients often.
Okay, that takes me to the next question, which is how important is a communication
skill for any AI engineer?
So that is one of the most important skills, because the data is messy in the real world.
And it is our communication, the way we interpret, the way we communicate will actually make
things a lot simpler.
So client expects only results, right?
And for that, we will need to walk through the journey.
If we are not able to communicate clearly, that is where miscommunications happen and
everything will be tricky.
So that is really important, AI can replace coding, but not the communication.
So we'll have to excel in that area.
I'm glad you mentioned communication.
And communication is also accompanied by many other skills like convincing people without
hurting their egos.
So do you see these scenarios where you are in a client meeting, there is a non-technical
stakeholder, business stakeholder, and they will do chat GPT and they will come up with
technical suggestions?
So that's a good point.
And it actually happened with me as well, where I was in one of the client meetings
and the product manager, he asked, the problem we are solving is very simple, it's based
on statistical and regex, we can get most of the tasks done.
And then when I explained him this approach, he just questioned me, where is AI in this?
So then I have to, you know, do some detailed breakdown, I will have, I know I did some
analysis, showing him the numbers, why we don't need that, what is the complexities
associated with that.
Then he understood, okay, let's stick to this approach.
So if you use regular expression, they don't like it.
Yeah, they want to have AI because they want to have LLMs in their architecture.
Okay, great.
So this is one of our meeting rooms, where, you know, key discussions will be happening.
And this is one thing that I like the most, whiteboard, because anything that actually
goes into implementation part, right, there'll be a lot of brainstorming happening, we'll
plan our architectures, and key discussions will be happening in this beautiful place.
That is really a good point.
If you don't have a right architecture, you might go on a wrong path altogether.
And then we realize, okay, something went wrong, and we'll have to, you know, yeah,
you waste a lot of time and money.
In this era of rapid technical advancement, what is one mindset shift that one needs to
have if they want to thrive in this new time?
So the very first thing I would say is unlearning a few things.
So I can say I am a developer, and I'm very good at Python.
And suddenly after 4.6 release, my skills have no value.
So I need to now start, how can I unlearn a few things?
How can I remove the baggage I'm carrying as a developer?
If we don't do that, AI will obviously replace me, and we don't add any value additional
to that.
So whatever the time that we are saving, giving the task to code or the AI, we'll have to
invest that time in the critical thinking, where we can add some value with the human
intelligence.
That cannot be replaced by AI.
So that is where we'll have to, you know, learn unlearning a thing.
Okay, that's, that's really a very good punchline.
We need to learn how to unlearn things.
Thank you Shikant for your time today and folks who are watching this video.
If you have any questions, please feel free to post in the comment box below.
Both me and Shikant will be ready to help you.
Our team has worked on more than 25 plus AI implementations for our clients based in US
and UAE.
And we can share our experience, share some tips based on the knowledge that we have gained
so far.
All right.
Thank you very much for watching.
Bye bye.
