My Journey to AI and Machine Learning as an Undergraduate Student

As I look back on my journey, I realize that it took me a long time to get started with Artificial Intelligence (AI) and Machine Learning (ML). After completing my first module, I realized that I needed to speed up my learning process. There were several challenges that came in the way, including family issues and personal struggles, which ultimately led to a three-month delay in my studies. However, I didn't let those setbacks deter me from pursuing my passion for AI and ML.

During that period, I had six months of study left, but I was determined to make the most of it. I spent every waking moment learning about AI and ML, pouring over notes, textbooks, and online resources. My goal was to get a solid understanding of the concepts and theories behind these technologies. I also wanted to be able to apply my knowledge in practical ways, so I spent hours working on projects and exercises that helped me develop my skills.

One of the key factors that helped me succeed was my willingness to learn mathematics deeply. As an undergraduate student, I had neglected math in the past, thinking it wasn't essential for my studies. However, I soon realized that mathematics is the foundation of AI and ML. It's the underlying language that enables us to understand and work with algorithms, data structures, and other technical concepts. Linear algebra, probability, statistics, and other mathematical topics all come into play in AI and ML, and I was determined to master them.

My journey also taught me the importance of focus and prioritization. As a student, it's easy to get distracted by extracurricular activities, social events, or part-time jobs. However, when you're pursuing a passion like AI and ML, it's essential to stay focused on your goals. I decided to take a job with Infosys, which helped me mitigate the pressure of finding a placement and allowed me to concentrate on my studies.

For undergraduate students who are interested in learning AI and ML as a fresh start, I would offer the following advice. First, don't wait until it's too late; start early. There's always something new to learn, and the sooner you begin, the better equipped you'll be. Spend more time on mathematics, even if it seems daunting at first. You'd be surprised at how many applications of math are present in AI and ML.

Next, be clear about what you want to achieve. What are your goals? Do you want to pursue a career in research or industry? Knowing what you want will help you focus all your energy on the right things. Don't spread yourself too thin by trying to learn everything at once; instead, concentrate on one or two areas that interest you the most.

Finally, don't be afraid to take calculated risks. Joining a startup or taking on a challenging project can be intimidating, but it's often the best way to gain experience and build your skills. In my case, I took a job with Infosys as a backup plan, which allowed me to focus on AI and ML while also having a safety net.

In conclusion, my journey to AI and ML has been a long and winding one, but it's taught me valuable lessons about the importance of persistence, focus, and hard work. If you're an undergraduate student who wants to learn AI and ML, I say don't wait – start now! With dedication and the right mindset, you can achieve your goals and set yourself up for a successful career in this exciting field.

"WEBVTTKind: captionsLanguage: enhi friends today we have sitaram who is one of our applied aa core students sitharam thank you very much for taking the time to share your learning journey which is fairly unique and interesting thank you very much thank you so much so thank you so much cool so let me give a little background so that everybody has the context then we'll go into your interview experiences and things like that so sitaram just joined as a data science intern at this company called z scalar or also spelled as z scalar which is a cloud information security company based out of silicon valley they're also publicly traded large company now i think they're about a 40 50 billion dollar company now so uh as part of his campus placement sitaram got placed for a full-time data scientist role like a fresher data scientist right out of college but the company also gave him the option to also do a six-month internship in his final semester so he's currently doing the internship on track to become a full-time employee when he come once he completes just a quick note z-scaler is a very interesting silicon valley startup i think a decade ago i think they started around 2006 2007 and they've grown massively so their compensations are comparable to your top product-based companies and sitaram is currently pursuing his b-tech while also doing his internship in computer science at graphical university so with that background sitaram let's get into your interview experiences your learning journeys because i hope this will help lot of undergraduate students realize that even as a fresher you can get into interesting data science roles at good product-based companies cool so let's get into your interview experiences right away so can you walk us through what interviews were there given that you're a refresher b tech computer science student what interviews were there what was the hardness let's go into the details of that yes so uh there were total of four rounds uh uh of which uh first and second round were like a first round was completely based on aptitude and reasoning and verbal uh so uh that was easy uh second round sorry sorry to cut you but what was the level of aptitude questions were they like cat level or were they easier what was the level of hardness uh no sir there were they were not that much hard if we know basic like uh they were based on number systems ratios percentages and graphs yes sir school math yes um so they were easy and some verbal questions like based on some vocabulary and all that uh after that uh second round was based on uh coding they gave they gave us a um coding question like um they gave us two option choice actually one was uh cnn classification and second was uh like hyper parameter tuning uh so that is a logistic regression uh problem so you can either choose machine learning or a different problem using the toolbox of your choice yes sir anyone uh we can solve uh either of those so data set correspondingly yes sir the data set was already loaded so they just wanted to check that do we know some coding related to data science or not so uh you have done i think all the machine learning assignments and i think you've done two or three deep learning assignments till now right yes sir so uh and there was also a choice question like uh for hyper parameter tuning hyper parameter tuning was sub rate one was logistic regression uh pure logistic regression without using any escalant uh libraries and all that oh so you have to implement a logistic regression from scratch oh that's exactly our assignment right yes sir that's exactly was our assignment so uh uh and they gave us were you also expected to implement a cnn from scratch because that's harder no sir uh they they just wanted us to uh the one data set was loaded like cat dog data set simple cat dog data set so images were already loaded and all that and all the processing was done we just were expected to uh write the code where we write like uh we initialize a model like sequential so you can use errors or pie torch and buildings libraries yes because implementing a cnn from scratch is much harder no sir we can use keras library for that so it was like uh 10 lines of code uh i guess so i i did all those three questions and uh because uh i already did the same exact assignment so the three questions were implement logistic regression for uh from scratch the second one and reader deep learning classifier using keras or pytorch third is a hyperparameter tuning yes sir third is like just optional and that was by hyper parameter tuning part you haven't started assignments anyway yes sir uh yes sir we we have done three assign three three questions in assignments as part of our assignment so i was uh they they were very easy and i did them so actually i was preparing for a mock interview as i was telling you before so yeah so i think you finished 15 assignments and you were revising for mock interviews yes interview happened okay and i i was like okay let's give us a let's give the real or actual test first so that i i will be ready and i never gave an interview or a test before uh even i haven't applied for software roles for whatever so so i didn't have uh any experience prior so i i was like okay i'm going to interview with uh applied ai so i have to be more rigorous so i applied for this role and i got selected in these two first rounds and then i got some confidence and i told my parents that uh okay the written part is now done now uh there will be interviews and in the first interview they asked about like uh the case studies which we have done in rsi uh as part of our course so i mentioned that uh we have done uh so uh one thing i've done is uh whenever you said like uh uh do uh end-to-end projects of our case study so i did that uh so uh i was very intrigued by uh taxi demand prediction problem you took the case study that we have explained and extended it and implemented it end-to-end from scratch yes sir for the for the taxi demand also i prepared a flag small flask api using class kpi a small website like thing and i deployed it in here and for question quora question pair also i i made a small website like to ask some question and uh uh i will show uh similar questions too so did you build these websites in streamlit or just using simple flask streamlets no sir uh for uh for uh taxi demand i use like uh flask api to simply build a very small app like thing for quora question pair and microsoft marvel my my malware assignment i did like i used streamlit uh to build a small uh one page website like thing perfect perfect and i deployed these three in heroku and i uploaded it in uh to github to showcase them like i've done this and that that's what they call uh caught eye of the interviewer and they asked me about uh my microsoft malware assignment in the first interview account because this company is also in information security right yes so uh they asked me whole about it like they started the interview and asked me about few personal questions like where are you from and all that after that they said okay go on with this project i want to hear about it in detail so i was like okay i i started and i explained them thoroughly like i also used dusk actually uh for microsoft malware assignment uh and because we are operating on large scale data and you can use multi-core and stuff like that and i was uh stubborn that i i will definitely run it on my laptop and whatever so i will not use google collab so for the malware detection assignment you implemented using dusk yes sir so because my system won't support that much use data so i had to use dusk or some other library very very good very good choice yeah so it took me some time like it took me one week to complete this assignment but it was completely worth it uh i was uh i can now implement any algorithm in task uh using task and support it to us so i understood what this i i came from a computer science background so i understand how distributed system works and all that i was introduced to hadoop and all systems also so it was very easy to pick up dusk for me so yeah uh that was and i reduced the time like uh you mentioned that uh in the notebook also like don't run this uh block if unless if it is necessary it will take 48 hours time to run so that i made uh two three hours i brought the time to three hours using the last and the distribution that dusk supports yes very good that was really rewarding and uh really uh motivating for me also that that helps you stand out in an interview right yes right you have taken initiative and you said if you just if you don't use any distributed like distribute system libraries like blast it takes about 48 hours and 95 percent of our students do that when they ask us we say use let's let's try to do it but most people don't put in the effort and the effort that you've put in paid to you enormously in interviews because you're showing initiative right to solve real problems yes very nice so and also uh like i wanted to uh use like uh i saw i go on through like kaggle forums as you said and talk solutions i went through top solutions and all that and i wanted to try all the solutions they tried so that gave me in-depth knowledge on what the problem was and how to tackle it and that was all the first rounds so they uh gone through the in this entire project in the first round and nothing else so that he was satisfied with my answer and he said okay well and i got selected to the next round so the final round was like uh uh the he asked me about two projects actually so these are also case studies which was uh graph data one is facebook friend recommendations system and second was uh like um taxi demand prediction problem because uh that is the only time series uh problem that i have done at that time and we have also improvised the solutions that we have discussed in the class productionized all these yes perfect yes sir so uh they've asked me about the those two projects and also they've uh asked me like uh uh what is the difference between uh batch normalization and uh sgd sorry uh uh yes batch has a batch uh gradient descent and uh stochastic gradient distance and uh what is the what is adam uh why why it is better than a delta rms prop and some other optimizers and some basic like uh deep learning questions and then he went on asking about these two projects in depth because and interviewer also said that they are working on a problem which is similar to facebook friend recommendation graph data so he wanted to know more about it and all the features are used and uh all that so i explained him thoroughly that all the features that we took features that we discussed earlier yes sir and that project that case study was totally focused on features and how how much features are important for us and how many features we can generate and what types of features we can incorporate and i also explained uh uh explain the uh matrix factorization uh features that you uh mentioned in the uh case study videos so he was intrigued by that and from there he went on towards uh unsupervised learning when i mentioned that i have also used like uh svd features and matrix factorization features he was like okay what is svd uh and why how is matrix factorization works uh what is matrix factorization and all that and uh they he also uh and in the second interview was a quite lengthy one uh it was it took me like one hour i guess so uh he also focused on unsupervised learning algorithms like clustering uh so he told me he asked me to explain k-means full full-on k-means algorithm so i was explaining uh as i was explaining taxi demand prediction i mentioned k means that we used uh claim k means to cluster the areas and all that so he was like okay uh these are all follow-up questions one more whenever you you said i've used a technique they would dive into it i think that's the strategy that they're employing yes sir very good so i mentioned k means so he was like okay what is k means and i say when i was explaining the uh taxi demand prediction i mentioned uh the same thing that you mentioned in the video explain while explaining it like uh k-means will result in uh evenly spare uh spread uh clusters uh globular shapes clusters and he was like why why do we why do why does k means result in this and is there any alternative for k means so that uh there uh there could be better initialization and all that so i mentioned okay there is one like k means plus plus and other algorithms and other extensions of k-means and then he was like uh what is okay this is all good but uh is there any other alternative for k-means like not gaming's extensions yes sir so if he wanted to test like uh if we want to cluster like based on density so what algorithm would you prefer so i then said that okay there is the an algorithm like db scan uh which we can use for that purpose and then he went on like okay what is davis can explain it to me so anything that you have said they wanted to test whether do you understand are you just blabbering these terms yes yes good good that's that's a typical sanity check for refresher to make sure that you know the basics of what you're saying that you've used or you claim you know yes that's it and then i explained these algorithms and he was satisfied and that was my very good very good very good so sitaram let's change tracks a little i also wanted to understand your learning journey right so what was your learning journey as an undergraduate student how did you take time out of your regular course work and college work to learn and i think you completed 18 or 19 assignments till now right yes so how did you prioritize both of them and how do you spend time here sir uh so uh i started early like uh in my sixth semester i guess yes uh sixth semester as my sixth semester started i enrolled in this course and that was the best thing happened to me literally uh i was going through some linkedin profile request that came to me so before accepting that and i found a workshop from uh that uh the person did from applied ai so i was like what is this and all that i was going through that uh and i went through went to your website website through that uh profile and i i gone through like uh syllabus and i was flattered by the syllabus and immediately i called uh one of the team and asked that is this really true is this what you're offering for us and they said yes uh do you are you interested and i was like i i asked my father i was already enrolled in uh specialization in machine learning and artificial intelligence in our university but uh um but they uh they were not teaching us properly so to be sure let's say probably they're teaching the basics like at undergraduate level it's hard for faculty members also to teach too much because they don't have time right yes sir so there was no math basically so i was not interested i am a math student and i'm i'm so i'm so much into mathematics so there was no mathematics so that uh was not really interested uh interesting to me at first but when i found your course the maths and the base you started up with like linear algebra probability and all that that took me like okay this is what i want but throughout your learning journey right how much time are you spending on daily basis or weekly basis both for assignments and the course content in general uh i spent a lot because uh i went through first uh uh i i kind of went like deep depth first search even though you mentioned like breath is important go on with breath but i i guess that's because of my age and maturity and that you have different learning strategies and curiosities we do see a lot of undergraduate students typically go depth first and then realize it's too deep come back and then yes that's that's what happened to me so first six months i spent like uh i watched through linear algebra of gilbert strang from mit course lectures yes sir the those were the best videos i was till today and after that uh calculus of uh from the same mit videos like i went through whenever i came across calculus in your course i complete them first and then i'll uh go into more advanced undergraduate level math or graduation yes sir because uh math was not taught to us that rigorously uh in computer science they just focused on some interpolation and all that but they haven't uh focused on calculus part that much yes and i i completed calculus rigorously like multi-variable calculus and all that first and that gave me very strong base base foundation and from then onwards uh this took me a long time and after that there were some problems due to covet in our family so three months were lost in that and after that uh again after six months work done there so i had i just had like six months and i just completed first module and then i realized okay now i get it okay so you had to speed up then yes sir i had to that's the challenge of depth first i had to complete this fast and there was no not really much time left so i was like okay i will spend uh all the time to this course and so because i can pick up any subject uh within one week before exams and i did that so i spent from morning to evening like uh i i used to sit and watch your videos and you had lesser amount of time to complete yes sir so that's why i didn't apply to any of the placements that were going on in my campus which were related to software roles so because i don't didn't wanted to waste time on that and application filling process was taking a long time like one hour two hours and i have to fill up the same things over and over again and got it and i know in my heart that i was not ready for those uh positions and uh i was not you were clear that you wanted to get into a data science machine yes sir so at that point i messaged you like sir this is this is what i'm feeling and i want i really want your advice so you replied me with a voice message that uh you first crack one or two software roles then uh you will get off that pressure off your head and then you can focus more then i i was like okay so i will apply for one job and i'll check and i applied for infosys and i got the job okay so then you said okay i have one job there i have one job mitigated yes sir i'm good so i focused entirely on uh this course and i gave a hundred percent like uh each day i used to spend eight hours at least on the course because yes for about four four months four five months four five months got it got it so i had another question for you which is what suggestions would you give to undergraduate students who are either in their second or third year right on how they can learn ai machine learning as a fresh as an undergraduate student right and prepare for careers as soon as they graduate what suggestions will you give to them or even to your one-year juniors right or even to your peers who are currently in the final year yes so if uh you are you are in uh second year or third year uh this is the high time for you to start uh because it is all you may think that there is a lot of time and all that but uh you will realize that there is a lot to cover to get uh to the point where the current research is going on and the current work is being done so please start early uh if you are in second and third year and start uh with uh maths uh don't neglect mathematics because uh that really opened many ideas to me that gave me total a different understanding of each algorithm and how it works and all that um it many of my peers thought that like uh maths why why are you studying that much depth maths it's not required but uh i have got i came across every every single math application in machine learning sir like every topic everything that we learn in our in our school days also like linear algebra probability statistics yes every single all maximum minimum everything comes into play here yes so that was really i think except complex numbers i think we use most of these concepts yes sir so that was really really interesting for me to see maths applications in this and i always wanted to be a researcher in future also so to join r d department of some company and do research work again so depth of mathematics is helpful whether you are in applied aspects of data science or in research aspects it will take you a long way i mean i mean more depth never hurts in the long run so that that is one advice i will give and spend more time and be clear what you want uh don't be like uh uh if you be clear at first only like what you want what your aim is then it will be easy to focus all your energy onto it then spread your energy to one or two things at the same time and also de-risk yourself have one software developer job that's what i think i always recommend even my audio replied to you i think that's what i said have one backup then focus all your energies on what you want yes perfect sitaram perfect thank you sitharam thank you very much for taking the time super happy to hear your successful journey keep digging deeper keep learning harder concepts because that will help you become a very good researcher in the long term so all the very best and thank you very much thank you so muchhi friends today we have sitaram who is one of our applied aa core students sitharam thank you very much for taking the time to share your learning journey which is fairly unique and interesting thank you very much thank you so much so thank you so much cool so let me give a little background so that everybody has the context then we'll go into your interview experiences and things like that so sitaram just joined as a data science intern at this company called z scalar or also spelled as z scalar which is a cloud information security company based out of silicon valley they're also publicly traded large company now i think they're about a 40 50 billion dollar company now so uh as part of his campus placement sitaram got placed for a full-time data scientist role like a fresher data scientist right out of college but the company also gave him the option to also do a six-month internship in his final semester so he's currently doing the internship on track to become a full-time employee when he come once he completes just a quick note z-scaler is a very interesting silicon valley startup i think a decade ago i think they started around 2006 2007 and they've grown massively so their compensations are comparable to your top product-based companies and sitaram is currently pursuing his b-tech while also doing his internship in computer science at graphical university so with that background sitaram let's get into your interview experiences your learning journeys because i hope this will help lot of undergraduate students realize that even as a fresher you can get into interesting data science roles at good product-based companies cool so let's get into your interview experiences right away so can you walk us through what interviews were there given that you're a refresher b tech computer science student what interviews were there what was the hardness let's go into the details of that yes so uh there were total of four rounds uh uh of which uh first and second round were like a first round was completely based on aptitude and reasoning and verbal uh so uh that was easy uh second round sorry sorry to cut you but what was the level of aptitude questions were they like cat level or were they easier what was the level of hardness uh no sir there were they were not that much hard if we know basic like uh they were based on number systems ratios percentages and graphs yes sir school math yes um so they were easy and some verbal questions like based on some vocabulary and all that uh after that uh second round was based on uh coding they gave they gave us a um coding question like um they gave us two option choice actually one was uh cnn classification and second was uh like hyper parameter tuning uh so that is a logistic regression uh problem so you can either choose machine learning or a different problem using the toolbox of your choice yes sir anyone uh we can solve uh either of those so data set correspondingly yes sir the data set was already loaded so they just wanted to check that do we know some coding related to data science or not so uh you have done i think all the machine learning assignments and i think you've done two or three deep learning assignments till now right yes sir so uh and there was also a choice question like uh for hyper parameter tuning hyper parameter tuning was sub rate one was logistic regression uh pure logistic regression without using any escalant uh libraries and all that oh so you have to implement a logistic regression from scratch oh that's exactly our assignment right yes sir that's exactly was our assignment so uh uh and they gave us were you also expected to implement a cnn from scratch because that's harder no sir uh they they just wanted us to uh the one data set was loaded like cat dog data set simple cat dog data set so images were already loaded and all that and all the processing was done we just were expected to uh write the code where we write like uh we initialize a model like sequential so you can use errors or pie torch and buildings libraries yes because implementing a cnn from scratch is much harder no sir we can use keras library for that so it was like uh 10 lines of code uh i guess so i i did all those three questions and uh because uh i already did the same exact assignment so the three questions were implement logistic regression for uh from scratch the second one and reader deep learning classifier using keras or pytorch third is a hyperparameter tuning yes sir third is like just optional and that was by hyper parameter tuning part you haven't started assignments anyway yes sir uh yes sir we we have done three assign three three questions in assignments as part of our assignment so i was uh they they were very easy and i did them so actually i was preparing for a mock interview as i was telling you before so yeah so i think you finished 15 assignments and you were revising for mock interviews yes interview happened okay and i i was like okay let's give us a let's give the real or actual test first so that i i will be ready and i never gave an interview or a test before uh even i haven't applied for software roles for whatever so so i didn't have uh any experience prior so i i was like okay i'm going to interview with uh applied ai so i have to be more rigorous so i applied for this role and i got selected in these two first rounds and then i got some confidence and i told my parents that uh okay the written part is now done now uh there will be interviews and in the first interview they asked about like uh the case studies which we have done in rsi uh as part of our course so i mentioned that uh we have done uh so uh one thing i've done is uh whenever you said like uh uh do uh end-to-end projects of our case study so i did that uh so uh i was very intrigued by uh taxi demand prediction problem you took the case study that we have explained and extended it and implemented it end-to-end from scratch yes sir for the for the taxi demand also i prepared a flag small flask api using class kpi a small website like thing and i deployed it in here and for question quora question pair also i i made a small website like to ask some question and uh uh i will show uh similar questions too so did you build these websites in streamlit or just using simple flask streamlets no sir uh for uh for uh taxi demand i use like uh flask api to simply build a very small app like thing for quora question pair and microsoft marvel my my malware assignment i did like i used streamlit uh to build a small uh one page website like thing perfect perfect and i deployed these three in heroku and i uploaded it in uh to github to showcase them like i've done this and that that's what they call uh caught eye of the interviewer and they asked me about uh my microsoft malware assignment in the first interview account because this company is also in information security right yes so uh they asked me whole about it like they started the interview and asked me about few personal questions like where are you from and all that after that they said okay go on with this project i want to hear about it in detail so i was like okay i i started and i explained them thoroughly like i also used dusk actually uh for microsoft malware assignment uh and because we are operating on large scale data and you can use multi-core and stuff like that and i was uh stubborn that i i will definitely run it on my laptop and whatever so i will not use google collab so for the malware detection assignment you implemented using dusk yes sir so because my system won't support that much use data so i had to use dusk or some other library very very good very good choice yeah so it took me some time like it took me one week to complete this assignment but it was completely worth it uh i was uh i can now implement any algorithm in task uh using task and support it to us so i understood what this i i came from a computer science background so i understand how distributed system works and all that i was introduced to hadoop and all systems also so it was very easy to pick up dusk for me so yeah uh that was and i reduced the time like uh you mentioned that uh in the notebook also like don't run this uh block if unless if it is necessary it will take 48 hours time to run so that i made uh two three hours i brought the time to three hours using the last and the distribution that dusk supports yes very good that was really rewarding and uh really uh motivating for me also that that helps you stand out in an interview right yes right you have taken initiative and you said if you just if you don't use any distributed like distribute system libraries like blast it takes about 48 hours and 95 percent of our students do that when they ask us we say use let's let's try to do it but most people don't put in the effort and the effort that you've put in paid to you enormously in interviews because you're showing initiative right to solve real problems yes very nice so and also uh like i wanted to uh use like uh i saw i go on through like kaggle forums as you said and talk solutions i went through top solutions and all that and i wanted to try all the solutions they tried so that gave me in-depth knowledge on what the problem was and how to tackle it and that was all the first rounds so they uh gone through the in this entire project in the first round and nothing else so that he was satisfied with my answer and he said okay well and i got selected to the next round so the final round was like uh uh the he asked me about two projects actually so these are also case studies which was uh graph data one is facebook friend recommendations system and second was uh like um taxi demand prediction problem because uh that is the only time series uh problem that i have done at that time and we have also improvised the solutions that we have discussed in the class productionized all these yes perfect yes sir so uh they've asked me about the those two projects and also they've uh asked me like uh uh what is the difference between uh batch normalization and uh sgd sorry uh uh yes batch has a batch uh gradient descent and uh stochastic gradient distance and uh what is the what is adam uh why why it is better than a delta rms prop and some other optimizers and some basic like uh deep learning questions and then he went on asking about these two projects in depth because and interviewer also said that they are working on a problem which is similar to facebook friend recommendation graph data so he wanted to know more about it and all the features are used and uh all that so i explained him thoroughly that all the features that we took features that we discussed earlier yes sir and that project that case study was totally focused on features and how how much features are important for us and how many features we can generate and what types of features we can incorporate and i also explained uh uh explain the uh matrix factorization uh features that you uh mentioned in the uh case study videos so he was intrigued by that and from there he went on towards uh unsupervised learning when i mentioned that i have also used like uh svd features and matrix factorization features he was like okay what is svd uh and why how is matrix factorization works uh what is matrix factorization and all that and uh they he also uh and in the second interview was a quite lengthy one uh it was it took me like one hour i guess so uh he also focused on unsupervised learning algorithms like clustering uh so he told me he asked me to explain k-means full full-on k-means algorithm so i was explaining uh as i was explaining taxi demand prediction i mentioned k means that we used uh claim k means to cluster the areas and all that so he was like okay uh these are all follow-up questions one more whenever you you said i've used a technique they would dive into it i think that's the strategy that they're employing yes sir very good so i mentioned k means so he was like okay what is k means and i say when i was explaining the uh taxi demand prediction i mentioned uh the same thing that you mentioned in the video explain while explaining it like uh k-means will result in uh evenly spare uh spread uh clusters uh globular shapes clusters and he was like why why do we why do why does k means result in this and is there any alternative for k means so that uh there uh there could be better initialization and all that so i mentioned okay there is one like k means plus plus and other algorithms and other extensions of k-means and then he was like uh what is okay this is all good but uh is there any other alternative for k-means like not gaming's extensions yes sir so if he wanted to test like uh if we want to cluster like based on density so what algorithm would you prefer so i then said that okay there is the an algorithm like db scan uh which we can use for that purpose and then he went on like okay what is davis can explain it to me so anything that you have said they wanted to test whether do you understand are you just blabbering these terms yes yes good good that's that's a typical sanity check for refresher to make sure that you know the basics of what you're saying that you've used or you claim you know yes that's it and then i explained these algorithms and he was satisfied and that was my very good very good very good so sitaram let's change tracks a little i also wanted to understand your learning journey right so what was your learning journey as an undergraduate student how did you take time out of your regular course work and college work to learn and i think you completed 18 or 19 assignments till now right yes so how did you prioritize both of them and how do you spend time here sir uh so uh i started early like uh in my sixth semester i guess yes uh sixth semester as my sixth semester started i enrolled in this course and that was the best thing happened to me literally uh i was going through some linkedin profile request that came to me so before accepting that and i found a workshop from uh that uh the person did from applied ai so i was like what is this and all that i was going through that uh and i went through went to your website website through that uh profile and i i gone through like uh syllabus and i was flattered by the syllabus and immediately i called uh one of the team and asked that is this really true is this what you're offering for us and they said yes uh do you are you interested and i was like i i asked my father i was already enrolled in uh specialization in machine learning and artificial intelligence in our university but uh um but they uh they were not teaching us properly so to be sure let's say probably they're teaching the basics like at undergraduate level it's hard for faculty members also to teach too much because they don't have time right yes sir so there was no math basically so i was not interested i am a math student and i'm i'm so i'm so much into mathematics so there was no mathematics so that uh was not really interested uh interesting to me at first but when i found your course the maths and the base you started up with like linear algebra probability and all that that took me like okay this is what i want but throughout your learning journey right how much time are you spending on daily basis or weekly basis both for assignments and the course content in general uh i spent a lot because uh i went through first uh uh i i kind of went like deep depth first search even though you mentioned like breath is important go on with breath but i i guess that's because of my age and maturity and that you have different learning strategies and curiosities we do see a lot of undergraduate students typically go depth first and then realize it's too deep come back and then yes that's that's what happened to me so first six months i spent like uh i watched through linear algebra of gilbert strang from mit course lectures yes sir the those were the best videos i was till today and after that uh calculus of uh from the same mit videos like i went through whenever i came across calculus in your course i complete them first and then i'll uh go into more advanced undergraduate level math or graduation yes sir because uh math was not taught to us that rigorously uh in computer science they just focused on some interpolation and all that but they haven't uh focused on calculus part that much yes and i i completed calculus rigorously like multi-variable calculus and all that first and that gave me very strong base base foundation and from then onwards uh this took me a long time and after that there were some problems due to covet in our family so three months were lost in that and after that uh again after six months work done there so i had i just had like six months and i just completed first module and then i realized okay now i get it okay so you had to speed up then yes sir i had to that's the challenge of depth first i had to complete this fast and there was no not really much time left so i was like okay i will spend uh all the time to this course and so because i can pick up any subject uh within one week before exams and i did that so i spent from morning to evening like uh i i used to sit and watch your videos and you had lesser amount of time to complete yes sir so that's why i didn't apply to any of the placements that were going on in my campus which were related to software roles so because i don't didn't wanted to waste time on that and application filling process was taking a long time like one hour two hours and i have to fill up the same things over and over again and got it and i know in my heart that i was not ready for those uh positions and uh i was not you were clear that you wanted to get into a data science machine yes sir so at that point i messaged you like sir this is this is what i'm feeling and i want i really want your advice so you replied me with a voice message that uh you first crack one or two software roles then uh you will get off that pressure off your head and then you can focus more then i i was like okay so i will apply for one job and i'll check and i applied for infosys and i got the job okay so then you said okay i have one job there i have one job mitigated yes sir i'm good so i focused entirely on uh this course and i gave a hundred percent like uh each day i used to spend eight hours at least on the course because yes for about four four months four five months four five months got it got it so i had another question for you which is what suggestions would you give to undergraduate students who are either in their second or third year right on how they can learn ai machine learning as a fresh as an undergraduate student right and prepare for careers as soon as they graduate what suggestions will you give to them or even to your one-year juniors right or even to your peers who are currently in the final year yes so if uh you are you are in uh second year or third year uh this is the high time for you to start uh because it is all you may think that there is a lot of time and all that but uh you will realize that there is a lot to cover to get uh to the point where the current research is going on and the current work is being done so please start early uh if you are in second and third year and start uh with uh maths uh don't neglect mathematics because uh that really opened many ideas to me that gave me total a different understanding of each algorithm and how it works and all that um it many of my peers thought that like uh maths why why are you studying that much depth maths it's not required but uh i have got i came across every every single math application in machine learning sir like every topic everything that we learn in our in our school days also like linear algebra probability statistics yes every single all maximum minimum everything comes into play here yes so that was really i think except complex numbers i think we use most of these concepts yes sir so that was really really interesting for me to see maths applications in this and i always wanted to be a researcher in future also so to join r d department of some company and do research work again so depth of mathematics is helpful whether you are in applied aspects of data science or in research aspects it will take you a long way i mean i mean more depth never hurts in the long run so that that is one advice i will give and spend more time and be clear what you want uh don't be like uh uh if you be clear at first only like what you want what your aim is then it will be easy to focus all your energy onto it then spread your energy to one or two things at the same time and also de-risk yourself have one software developer job that's what i think i always recommend even my audio replied to you i think that's what i said have one backup then focus all your energies on what you want yes perfect sitaram perfect thank you sitharam thank you very much for taking the time super happy to hear your successful journey keep digging deeper keep learning harder concepts because that will help you become a very good researcher in the long term so all the very best and thank you very much thank you so much\n"