Student Name: Koustav Dalui

Current Course: MSQE

Degree Before Masters: BA Economics

Company Name: InMobi
Job Role: ML Engineer
Duration of the Job Role: Full time

Hiring Process Overview:
Apti > 2 technical > 1 behavioural > HR

Questions Asked in Online Aptitude Test (if applicable):
Around 25 qs on ML and DL and 1 coding qs of sliding window. It was a 90min tesr

Number of Interview Rounds:
More than 2

Questions Asked in Round 1:
After basic introduction, i was asked to pick any of my projects and explain. From there the interviewer tangentially went into other topics.

I was asked to derive beta under both lasso and ridge. Why lasso is used for feature selection. How ridge helps with multicollinearity. Explain the assumptions of linear reg and find estimations using MLE.

Next they asked to explain the various types of gradient descent and how they differ. I was asked to prove that the gradient under SGD is an unbaised estimator of the gradient under BGD.

Next they went to Boosting algorithm. Given n-1 trees in a gradient boost , explain the process how the next tree is created. How Xgboost differ from gradient boosting and write down the objective function to minimize for both. Why use boosting over decision trees. And how boosting reduces overfitting.

At last there was a coding question : Given an array, return True if any number occurs more than once. They expect the interviewee to write the proper code and also explain why that is an efficient solution.

Questions Asked in Round 2:
This round started with a few simple probability questions: If a die is rolled thrice, what is probability of getting 2 consecutive 3s Next the interviewer went into coding:

  1. If there r n people in a town. And we r given each person’s birth and death year. In which year is the most number of people alive.

  2. Print : abcd, bcd, cd, d . Using a recursive function.

  3. Given a set of parantheses. Return True if they r valid and balanced.

  4. Given a set of equation in an array eg. [ a==b, b==c, c!=a, d==a]. Return if this is a valid set of equations or not.

Mostly all were leetcode questions

Questions Asked in Further Rounds:
In the behavioural round. They just ask to pick any project and explain. This is a short round of 20-30 mins. Try to pick a project that is relevant to the kind of work they do. A ML or DL related project will be helpful rather than some econometric or time series related ones.

Job Role Experience:

Suggestions for Candidates (including resources, books, websites):
Practice coding. They r a tech company and expect high proficiency in coding.

Study up the internal math of any algorithm mentioned in CV.

Additional Resources:

Previous Job Role (if applicable):