Mock Interview Prep Material

Notes:

• Do not attribute a question to a specific company.
• If a code review is related to a question, try not to give it to someone who has done one or the other (e.g. someone who has recently done list implementation for a class or an interview question shouldn't do that for code review).
• People should write real code and not pseudo-code. If their code is lacking, they need feedback. (Nice, polite feedback of course.)
  • Flip side, upon joining a company, be prepared for your co-workers or managers to constructively let you know you did not approach a problem correctly. So try to get used to it.
• "I don't know" is not the end of the world!! Especially not in mocks. Give as few hints as possible but definitely hint to move them forward.
• Questions should be limited to 20-50 minutes depending on the question and interviewee. If they cannot solve it, then need to study the topic more or use easier questions.
• They shouldn't need to look anything up for any question. If there are things they would Google, they should ask the interviewer who will answer at their discression.

TODO:

• Collection of interview questions to be sourced from:

  • Books:
    • Cracking the coding interview (you can google the 4th edition pdf, tons of great questions.)
  • Websites:
    • http://projecteuler.net/ (not all of them, but all the more reason to source well)
    • http://codeeval.com/
    • rapleaf (or liveramp) challenge questions (e.g. "array absurdity")
    • facebook challenges
    • https://www.hackerrank.com/
    • http://poj.org/
    • http://www.glassdoor.com/index.htm

• Code review for C, C++, Java, Python (with errors ranked by importance)

  • All code should compile
  • Code reviews last 30-50 minutes
  • For many people, reading code is new and hard
  • An easy bit of broken-code to write is an ArrayList in Java with a memory leak, or a LinkedList that doesn't handle its tail pointer correctly.
  • Common problems include null errors, and in Java, .equals() vs. ==

• Rank by difficulty and class pre-reqs (ECS 40, 60, 122A)

• Question classes

  • OOP | inheritance vs composition | polymorphism | encapsulation
  • Data Structures | trees | binary | searching | balance | graphs | BFS | DFS | A* | sets | maps (aka hashtables) | sizing | hash function | runtime (hint: O(1))
  • Algorithms | recursion | recursion |recursion | dynamic programming

What not to do

• Use a language you don't know.
  • For example, if you wish to use Python, is slicing a constant or O(n) operation? Does it depend on what is being sliced? If you're not sure of the answers to these questions, you shouldn't be using Python in an interview.

Do do

• Focus on scale (big-O).
• Name big-Os on all operations, on both time and space!
• Recursion is almost always at least O(log n) space! (unless tail recursive).

Questions:

• (Open form) You are downloading two files. One reports it will be completed in 4 minutes, the other in 8. When will both be done? The interviewee should ask questions about everything, trying to make as few assumptions as possible.
  • Are both on the same server?
  • Are they using the same amount of bandwidth?
  • Is the server limiting bandwidth per connection, or per client?
• Create a data structure which supports two operations: add() to insert an integer, and test() which accepts an integer and returns True if two integers have been add()'d such that their sum is equal to the value being tested.
• Minimize dot product (Mikel's jelly beans)
• Create a data structure which contains integers and support constant add(), remove(), and getRandom() operations.
• List implementation. Examples include ArrayList and LinkedList. Test for encapsulation and polymorphism.
• You have a commercial garage. Describe what kind of Objects are in it, and how they are related to each other with regard to inheritance and composition.
  • cars
  • tires
  • attendants
  • engines
• A data structure, such as a Python list or a Java Collection may support a .sort() method. After this method has been performed, when no other operations have been performed on the list, imagine a user calls list.index() in Python or an equivalent on the Java Collection. Can that operation be optimized to be better than O(n)? If so, how? If not, why not?
  • Answer is no, interviewee should elaborate.
  • Follow up question: are there languages that don't have this same problem? (Yes: Lisp, Prolog, etc.)
• Java question: can Java leak memory? If so, how? If not, why not?
  • Yes. Interviewee should elaborate.
• You are given a linked list. You must select a random element from it, and may make at most one pass over it.
  • Doable in O(n) time and O(1) space.
  • Make sure the result is truly random.
• Implement square root.
  • Not as mathy as it may seem. Solution involves guessing and checking iteratively.
• Have you ever lead a team? How big?
  • Ask what each member did. How the interviewee helped them, who didn't need help. Challenges to leadership.
  • If it turns out they don't know who did what, or they claimed to lead a team that was large but when probed say "so and so didn't really do anything" question whether they should consider them part of the team they lead.
• Interviewer should pick some product of a big company, e.g. Amazon, Facebook, Google, LinkedIn, and discuss implementation of that product. What would it take to replicate it and produce a version 1.0? What would be in the 2.0?
• Design and implement a Map which respects the order in which items are added.
  • One solution involves internally using a HashMap and a DoublyLinkedList.
• I (Michael) should try to remember my Factset phone screen question(s).

How to succeed:

• Make the interviewer personally want you on their team.
• Be nice!
• Think outloud as much as possible.
• If you think of a "dumb, easy" way to do something, state that outloud, and then say why you'd choose not to do it that way. This shows that you're thoughtful.
• Enthusiasm is a good thing.
• For loops over while -- http://nedbatchelder.com/text/iter.html