I was interviewed at One Horizon Centre, Gurgaon Office. It was a pool drive for EDA (Enterprise Digital Analytics) team. Around 20 to 30 pre-shortlisted candidates were present over there. Process had 2 rounds, First round was to check your technical skill set & second round was for checking behavioural & fitness test. My first round went really fine ( or may be it's just my perception ). The interviewers ( Varuni & Sakshi ) were really good, it was a fun discussion. Questions were mainly based on my past experience. I began with introducing myself & explained my project on demand & supply prediction for fleet management of ride sharing and ride hailing companies. Some of the questions asked were: 1. Why did you use SARIMA & why not other regression techniques like Random forest & GBTs ? My Answer: My data had seasonality & I had to do something for damped regression. Also, RF & GBTs were giving me poor MSE . 2. For one of the project why didn't you use Logistic Regression in place of RF? My Answer: I started by saying Logistic Regression is another classification technique but categorical variables have to be converted to numerical & issues like cross validation and dealing with collinearity are handled automatically with RF. 3. Difference between Bagging & Boosting? Bagging : Bootstrapped Aggregating Boosting: Bootstrapped Weighted aggregating Both are used to combine weak learners into a strong learner with different averaging & input data selection technique 4. for recommender system why did you use euclidean distance with Nearest Neighbors? My answer: I could have used Manhattan distance or cosine similarity too but if something is already working for me why waste time on something else. One case study: 5. How will you use AmEx data to find out who is going to buy a car? I said I will use age, gender, job designation & CTC as some features for this problem but they insisted me to think about something else. Nothing came to my mind at that time & eventually we moved ahead with other questions Two Guess estimates were also there: 6. How many planes are their in the sky right now? My answer was horrible but there's no right or wrong for Guess Estimates 7. How many footballs are there in Delhi? I made a very bad assumption that each household has one kid & every kid has a football & every 3rd adult also have one. There are x household per sq km and delhi's area is 1400 sq km so answer would be 1400*x(1+1/3 ). My Second round was not behavioural round , it was again a technical round, it was okayish it wasn't bad either. It was taken by a guy named Mrigank, I later on found out he's one of the director of Data Sciences at AmEx. He tried grilling me down on RF & Decision Trees too & some of the projects were discussed too. Some of the questions were: 1. How decision trees are different from Random forest? Me: RF is ensemble of DTs 2. Why node splits at a different feature in DTs & RF? Me: entropy changes at each DT because input is different 3. What happens when all DTs have given their output in RF? Me: Averaging, Voting or you can do bagging , boosting 4. What was Y & x in your problem? Me: Y are the classes, x is my dataset 5. What are SVMs? Explain them 6. What is the role of dimensions in SVM? 7. Some other 3, 4 questions regarding ML, 3 4 other from my project 8. One classical interview puzzle - Out of 25 horses find fastest 3 (I honestly told him I already know the answer) 9. One case study: How will you use AmEx data to help Chayos improve their business? My answer: I will try to extract as many features possible from the data like age group, gender, industry, timings , gave reasons too why I chose them and then I will use sklearn.SelectKBest to find best features to help me in modeling. The key is to elaborate your answer try to add as many important features. By the end of the discussion he asked me 10. what's your preference data science or strategy? Me : Data Science but I do not have any issue with Strategy too. Result: Rejected Interviews are 80% luck, 20% your knowledge(Pareto's rule) . They might have selected Someone more experienced & luckier than me. I also lacked experience just 2.6 years whereas minimum asked was 3 years. All other candidates were more qualified they had Master's degree too. My suggestion to other candidates: Please ask for Referrals, if you have a friend at AmEx that really helps. CTC that you asked could be way higher than what others are asking, be cautious. Read Random forests , and all kind of trees XD.