LinkedIn Data Scientist - Inference & Algorithms interview questions

1. Coding (25 mins) - I was asked; Given an array, find a number (or rather it’s index) to start with such that the same is never negative. Similar to “ 1413. Minimum Value to Get Positive Step by Step Sum” on leetcode 2. Machine learning techniques (10 mins) - I was asked about how to improve LinkedIn’s premium subscription among users. This goes under Logistic Regression category. So the second part of the question was knowing that 1% of users are subscribed, how do we run a model with such imbalance (100 subscribers vs 9900 free users)? Well make it balanced, sample a 100 non-subscriber data and consider former subscribers. Also, consider that not all 100 current subscriber data is clean. 3. Statistics - I was asked about A/B testing for email marketing. It started out as simple as comparing percentage of success rates between email A and B. But then advanced to asking how to utilize statistical significance and statistical power in comparing these success rates. If I don’t move on to the next round, it is because of this Stats section.