What Is Asked In Interviews For Data Science With Genertaive AI Roles?

hello all my name is krishak and welcome to my YouTube channel so guys there is some something going wrong with my camera so uh I just made some another setup so that I wanted to upload a video because from past 3 days I did not upload a video uh this video that I am going to probably upload is super important for everyone out there who are looking to crack data science interviews or generative AI for generative AI roles because recently I had a discussion with one of my student who have recently cleared the interview and he has been hired as a generative AI engineer and he was somewhere around 6 months of experience that also only with internship so I will talk about what all important questions were specifically asked with respect to the interview because many people still have that confusion Kish if we are applying for the generative AI role what kind of questions we can expect from the interviewers whether data science will be covered whether the machine learning deep learning will be covered so uh uh the student uh who has actually cleared had had a detailed interview rounds they were around three to four rounds and I will be covering what all things were specifically asked and just to give you a brief idea about it like if you also preparing you should definitely prepare in this manner so that it'll be beneficial for the interview so let me quickly go ahead and share my screen over here in front of you so uh generative AI interview I have written I've I've listed down some of the important topics the main main models topics that was specific Ally asked in the interview itself okay so initially um there was a round with respect to Python and uh the kind of questions that were asked with respect to the python were from basic to intermediate not also Advanced basic to intermediate were sufficient over here with respect to python now um there was a use case that was given to the student to the candidate basically and uh he was also given two days time to complete the task with the help of python so he was able B to complete it because the task was pretty much easier uh what kind of task was actually given um he was not allowed to disclose that so uh but he just said that he was given a task uh for which he was given two days time and he actually give the task to them okay now going forward uh as he went probably you know the next round that he had was with respect to interview now overall uh the kind of topics that were covered were like statistics machine learning okay uh deep learning and open source llms okay and I will probably deep dive into each and everything that was basically asked right and they were also questions with respect to paid llms and databases okay now you will be seeing over here the majority of the portions and since this was a role for generative AI engineer and this was with respect to llm models okay not Li models it was with respect to large language models so that the kind of task that you will be able to solve is with respect to NLP that is natural language processing that basically means the business use cases that you are solving this will be something related to text okay and that is the reason we say it as natural language processing now with respect to python as I already said basic to intermediate was there uh over here and then when we go the person actually started the interviewer started to ask questions with respect to statistics now in statistics there are two main parts you know one is the descriptive statistics and one is the inferential statistics uh with respect to the candidate the recruiter specifically focused on inferential statistics now when we talk about inferential statistics that is something related to hypothesis testing and then in hypothesis testing various test let's say I've given some example like Z test T Test Ki Square test right Kai Square test um there was a test with respect to Anova test right so I'm listing down all the tests that are available in the hypothesis testing but the person specifically got asked question with respect to Kai square and Anova okay and the main outcome of this part of this module was to understand how this test are used in real world scenarios okay real world scenarios so this was the main thing over there right the reason why I'm writing all the other test because you need to prepare in this specific way itself anyhow what I will do is that all the playlist link I will be giving in the description of this particular video which you can go and refer because everything that I think that I have actually uploaded in my YouTube channel you don't even have to go anywhere you know you just have to refer them all the materials all the questions how it is related to a real world scenario everything is given so with the help of this preparation I think you'll be able to clear it very much easily because this is the feedback that I got from the person who has already cleared it okay now since as I said the majority of the task was with respect to llm models in machine learning much algorithm point of things was not asked but something task related to NLP like one of the questions were asked with respect to text embedding okay text embeddings now text embeddings there are different different techniques like tfidf bag of words you know but the person when he was taking the interview they specifically asked about word to W and the question was asked like how word to W is trained from scratch okay how how it is basically trained right so this was the question that was asked over here and again when I talk about word to W I'm just including it in machine learning but the way that how word to W models are basically trained there specifically comes in deep learning part also right because over here also use use an artificial neural network to train this particular word to models okay so the question was asked deep dived in word to and uh uh there was also some examples where they had asked about the implementation part also okay like how the data set will be that if you want to train a word to model from scratch how do you think you're going to prepare your data set what will be the input values what will be the output values what should be the vector size you know all this kind of questions were specifically asked you know and uh there were also some of the mathematical Concepts that were asked right like what is cosine similarity so these all are something very important you should know it right so similarity what is cosine similarity what is similarity score so everything was basically covered right some amount of mathematical questions along with machine learning algor uh and machine learning uh questions with respect to NLP use cases okay uh and he also was uh he also said that they had asked some of the question with respect to simple linear regression to understand his Basics okay simple linear regression okay so at least this algorithm I think you should be really really good good at now with respect to deep learning NLP over here two important questions were asked but at the end of the day what all important questions do you think over here are like activation functions right activation function loss functions then you specifically have optimizers optimizers are very important okay and uh if I probably talk with respect to the interview this was the main question that was asked and with respect to all the neural network like Ann CNN and all if I consider there was one neural network that was basically asked that is basically Transformer and Bert okay and Bert so this two like 30% of the entire interview were focused on this Transformers and B you know and how is the architecture what is encoder decoder what is attention what is self attention how can you implement it from the scratch how can you use hugging face and implement this every everything was asked in this so the majority of the time right 30% of the entire interview time was spent on this Transformer and but and already again as I said in my YouTube channel I already uploaded the video all you have to do is that go ahead and refer it completely in detailed it is a 2hour session step by step how things work everything has been explained in that particular video okay so it is up to you please go ahead and check it out okay now the reason what I feel is that Transformers is asked because right now with respect to all the models that you see all the large language models you know the base is Transformers right so Transformers really becomes a very important topic when I say Transformer there are multiple topics that comes in mind okay so let me write down all those topics right one is attention is all you need attention is all you need our super important research paper and this interview that was happened just a week back okay and and uh just a week back also I had uploaded a video a research video uh a research paper had come something related to one bit llm right and just imagine this question was also asked in the interview okay because the the the candidate was very much curious to talk about this you know so uh trust me this is the best feeling that I get you know whenever I try to upload a video and suddenly it is asked in the interview that gives me the best ever feeling now with respect to deep learning NLP as I said attention is all you need what is encoder decoder activation function lost function two important question as I said optimizers were asked to him and Transformer and Bert was asked right then they started focusing on open source llms right they started talking about Lama 2 right Google Gamma model right Google Gamma model so how this how these models are specifically trained so if I probably consider with respect to Lama 2 how this model is trained again I've explained in depth in my video that is the best feeling I get guys okay and the major question is that when should you probably go ahead with open source or paid right so I have also created this video also right it completely depends on the use case data security many more things right now with respect to paid llms uh again the majority of the question was asked in open AI you know and uh the other models that are there like cloudy cloudy 3 has recently come right and if I probably talk about this architecture that is serverless serverless from Amazon Bedrock right so Amazon Bedrock what it has done it has given you as serverless apis to exess all this kind of llm models different different llm models uh you have a lot of different different llm models even from Amazon you have models like Titan you have Lama 2 open source model different different models itself how you can actually use serverless Amazon Bedrock because the cost is also very very much less you don't even have to worry about the infrastructure and all and with respect to this when they were asking with respect to the paid llm models the major thing that they focused on was on Frameworks how you can with which Frameworks you can specifically work like Lang chain right Lang chain then you have llama llama Index right llama index how you can work with this what are the functionalities between langin versus what is the differences between langin versus Lama index there is also one another framework which is called a chainlet right so all this kind of Frameworks were specifically asked what are additional functionalities you have in this Frameworks you know if I talk about databases then there were questions asked with respect to Vector database right there were questions asked with respect to my SQL and a no SQL database I'd again suggest that please prepare for these three kind of databases because Vector database is a no SQL database itself my SQL database and no SQL database how a vector database is specifically used in in creating lot of different different applications when these all questions were specifically asked trust me when I say I have written overall the topics that were focused on but all these questions that were asked were based on the projects right any projects that you implement with respect to gen Ai and I tell you guys right now whatever projects that you create just by your own understand those projects are also super important because right now geni is still developing people are developing more amazing things probably in upcoming two to 3 years you know there will be many things that will be coming up when we talk about projects generative AI I've since I've also uploaded videos with respect to llm Ops okay llm Ops if you probably see my playlist over here right in my in my channel itself right now this project while they were explaining this the candidate was explaining he covered almost each and everything over here right initially from here to uh if I talk about this deep learning part you know normal questions were asked and then when the person when the interviewer went for the project part he started explaining about all these paid llm models that were used the Frameworks that were used the databases that were used why we did not go ahead with this database what are the deployment mechanism that was specifically asked now in langin there were simple amazing questions that were asked with respect to Lang serve right there were question asked with respect to Lang Smith because these are some new features that are coming with respect to deployment and creating apis a simple and easy way right and he was also able to talk about this so if I talk about the entire interview you can probably see that still if I say 20% of the interview they have still focused on if I probably talk about 20% of the interview they have still focused on basic concepts right and then if I talk about deep learning 30% was actually covered right deep learning part specifically Transformers and all if I talk about project part right another 30 percentage was covered so 30 20 30 is somewhere around 80% remaining 20% will be on some of the other things right so this way the entire interview had actually gone right so you also need to prepare in this specific way the best part is that I've created all this kind of playlist now right now in LinkedIn also if you go ahead and see every day I'm trying to post a transition story I have lot of transition story still I have not posted it out it is up to you go ahead and probably check it out that will definitely give you the motivation so I hope you like this particular video this was it for my side I'll see you in the next video have a great day thank you and all take care bye-bye

Channel: Krish Naik

Views: 25,664

Keywords: yt:cc=on, generative ai interviews, data science interviews, machine learning interviews, llm end to end projects, statistics tutorials

Id: UifWm9h96Ec

Length: 14min 31sec (871 seconds)

Published: Fri Mar 08 2024