From the online application (Bloomberg website) to the phone interview took around 4 weeks. A first interview was scheduled within 2 weeks, but the interviewer didn't call, so we rescheduled.

The phone interview was quite straightforward, and I was asked the following questions:

1. Generate all 3-digit combinations of 1-9, but not repeats in the sense that if you have 123, then 231, 321, etc., are not allowed. (nested for loops)
2. Reverse a given sentence (no use of library methods, arbitrary spaces between words). Make it faster. (start at the end and find and print words one-by-one)
3. You are given a 3L and a 5L container and an unlimited supply of water. How to make 1L? (I found a longer solution, but the short one is pour 3L into 5 twice.)
4. A program/algorithm that, given mL and nL buckets, figures out how (if possible) to make kL. (We discussed the solution, and the interviewer was nice and gave me a hint. It's a DP problem.)

The next day, I received an invitation for an on-site interview. The travel schedule part was straightforward and convenient.

I was invited to the NY office. There was a group of around 30 people like me. They hold you for a few minutes to issue a security badge. Then there is a tour about all the 31 or so aquariums and fishes at Bloomberg. Eventually, there is breakfast, and then the interviews come find you. There are 4 rounds of interviews. Rounds 1-2 are technical, 3 is management, and 4 is HR.

Some questions I remember:

1. Find the shortest path on a chessboard from the lower-left corner to the top-right. You can move right and up only, and some cells are inaccessible. (DFS)
2. You have sorted arrays A and B with sizes m and n. Assume at the end of array "A" you have "n" empty slots. How to merge with a minimum number of operations? (start at the end instead of the beginning). Prove it works (induction).
3. Some runners running in a field with some "k" sensors. We get events of running locations. How to print the top "m" runners? (array of lists or hash table of lists depending on the location of the sensors; each list is one sensor). Some questions on how to identify the ordering within each list.
4. CPU Cache implementation. How to implement the least frequently used procedure.
5. Scheduling some jobs over "m" servers. Centralized vs. distributed implementation? How to implement.

The management interview was interesting, and I liked the two managers. HR was fine too. I did not do any negotiations.

The chessboard problem.

I first mentioned BFS and the use of a Queue, but the interviewer kept asking about a data structure with less memory that I could extract due to the simple structure of the graph. I did not understand what he meant.

I eventually mentioned DFS and proved it works and uses less memory.

The cache problem took a long time. I felt I had failed it at some point. I had no prior knowledge of the topic since I am not a CS major.

I eventually used an array to store the access time to different items and an O(n) search through it to find the least frequently used one. The interviewer did not raise the complexity. He wanted me to write code on paper (which is hard, especially in C).