Source — Netflix Gifs

Thought it would be fun. I was wrong(Data Science Interview)

As promised to you all, this is a story of the time when I just started my journey in the Data Science field, though not that the journey is finished yet, and this was not the first time I gave an interview but this was surely the first time to gain experience and not to change my job. This blog is to just give my fellow “Aspirers” how a Data Science interview is, please don’t dwell on the answers much ( I was not able to answer many :)).

Meetings on Zoom or telephone is now the new normal but for interviews and that too first in the process, this was always the way to go.

I: So Nirmal what are the projects you have done in the data science field?

M: “Ohh, he has not seen my resume yet” Actually I don’t have any practical experience in this field, I have been a backend developer in Java professionally, and all the things related to the Data Science field I did are in terms of Academic projects.

I: Ok, so let’s go through one of your academic projects then.

M: Yeah sure, blah blah blah.. telling everything about one of my academic projects which I was doing in my PG Diploma course, he sure was convinced with the fact that I was doing the project myself. (which I literally was :))

I thought I started well, but the story has just begun. I had no idea that this was going to be a rapid-fire round of the tech questions and would turn into a grilling session for me.

I: How many features were there in the model?

M: Including the internal and external data there were around 100 features out of which around 30 were derived.

I: What is the statistical measure you take to separate out the significant and the insignificant features?

M: We generally take the p-value into consideration after building the first model and then remove the features with a high p-value in the further models, and we also remove the correlated features using VIF analysis.

I: What is the test to see the correlation between categorical-categorical and categorical-quantitative.

M: Though I know the names like Chi-Square and ANOVA tests but it did not strike me at that point in time and I said that generally I have only done using visualizations and remove the variables which are collinear.

I: What is the normal distribution and what statistics do you see to identify it without seeing it visually.

M: Concept of median, mode and mean equal and 50 percent of the variables lying on either side of the mean and 1:2:3 rule. (I felt confident after this answer)

I: What is class imbalance and what is the problem with that?

M: Answered perfectly.

I: Which evaluation metric out of Precision, Recall will it decrease?

M: Totally confused, remembering the formulas of both and trying to get the answer.

I: Okay never mind, let’s move ahead, what is the Odds Ratio in Logistic Regression?

M: Answered but not with confidence.

I: What are different kernels in the SVM?

M: “Shit” I know SVM but no idea right now.

I: Okay, what are the measures on which the decision tree split in case of regression problems?

M: “I know the answer to classification problems but not for the regression” No idea after thinking 30 secs.

The above were the exact thoughts in my mind.

I: What is bagging and OOB error in the random forest?

M: “Took a sigh in my mind, I knew this answer” Explained thoroughly.

I: Difference between bagging and boosting?

M: Explained

I: Let’s move on to Neural Networks, what are global and local minima?

M: Had the mathematical intuition, answered through that.

I: What are activation functions? and the difference between Relu and Sigmoid?

M: This was the first time I was hearing the above term.

I: What are vanishing and exploding gradient concepts?

M: Ummm.. long silence, “Neural Networks was not reached as a part of my course yet to be fair.”:)

I: What are optimizers and the differences between Adma and Adagrad.

M: Umm.. sorry I don’t have enough idea about Neural Networks.

I: You should be knowing NLP right?

M: No, have never used it yet.

I: Okay, I am done, you have any questions for me?

M: “Mind-wandering somewhere” No sir, thanks.

This was the first 25-minute interview experience for me in this field.

So, guys, I think you know the result of the interview so I won’t be telling you that. :)

Share your thoughts in the comment section and of course, do share your experiences.

It does not matter on which side of the table you are, what matters is your story.

Link to my other article

https://commondatascientist.medium.com/statistics-in-data-science-part-1-b39e89db9e4b