Deloitte visited IIT this placement season to hire consultants in Data Sciences Profile.
Round 1
Firstly, an online test was conducted(120 minutes duration) consisting of questions on Quantitative (30), Aptitude (30), Verbal (30) and Probability and Statistics (30). If you have prepared well, this won’t be very difficult. However, make sure you study on topics like Hypothesis testing, ANOVA, and statistical Inference for cracking the last section, in which the focus was on statistics.
Out of 900 students, 35 were shortlisted and they kept a waitlist of 15 students. Then, we had an online PPT in which the interviewers walked us through the various stages of interview
Next, There were 3 rounds of interviews.
HR Interview
My first round of interview was HR only and it was taken by the Director of the Analytics team and Actuaries team and it was more like a conversation. They first introduced themselves and asked me to do the same. Then, they asked me to walk them through the CV.
- What did you do in your third year intern? What is the business aspect of this?
- Explain a little in detail about the different solutions that you went through for this and also tell us why you rejected them?
- What do you want to do in future? Managerial or technical? I told him that I want to be a Data Scientist and not a Data Analyst, he asked me the difference
- I asked him does he still like to do what he does after spending such a long time in the industry? How do priorities change after 15 years?
Technical Interview
The interviewer was a very nice guy and he asked me how can I look so fresh during the placements and I cracked a joke about being naturally good looking.
- Took a look at my CV and asked what a regression forest it? How do you implement it?
- Asked about what kind of data sets we used and how did we obtain them?
- What was the main problem with this data set and how did we overcome it? I answered Dimensionality reduction and walked him through the series of steps I took.
- Given a data set where all the data points look negatively correlated, but actually are sets of points from different sources, which are positively correlated. How do you account for this?
- I told him that’s the basis of regression forest. On IID assumption of regression, how to bring stationarity in time series?
- What other ways to bring stationarity? I answered taking log, first order difference, percentage etc.
- To launch a product into the market, which data sets would you like to pull in? Which markets to be analysed?
- On product cannibalization, why did Apple launch iPad to remove iPod?
- On self cannibalization of product.
Technical Interview -2
He was an IIT alum from my department.
- Difference between OLS and MLE.
- Different metrics used for estimating the size of weights.
- Series of steps for dimensionality reduction. I answered the same as before. But then he asked me to come up with a completely different approach for this. I said plotting graphs on artificial Neural networks, how do we find weights, why do we add bias.
- A little talk on backpropogation.
- Is hard-limit function of perceptron, implementable, why ? why not?
- What is the benefit of using a sigmoid function for hard-limit function
- Why don’t we use a piecewise slant line if we want to adjust threshold, why use a sigmoid?
- Derive the weights of simple linear regression using one variable only
- Asked preferred language. Write code for implementing a change in size of a matrix in R. I used a subset, apply with a for loop. He used a dplyr package.
- For preparation, take online courses from Coursera (Andrew NG, ML) and edX (Analytics edge). Follow AnalyticsVidhya and r-bloggers and follow their learning curve page.
Finally, they selected 4 people out of the 35 and I was one of them. Last tip, although, Deloitte didn’t ask for algorithms, most companies do. So, do some coding and algorithms from geekforgeeks and Hackerrank. Overall, it was a good experience, and a memorable one too!