Target: Oracle, senrior member of technical staff
Four 1 hour rounds, each with technical + behavioral + some design.

Round 1

Tell me about a time when you were unable to meet deadlines.

Given a postfix expression (input: vector of strings), covert it into an infix expression(output: one string) with no redundant braces.

Example

Input: ["12", "11", "y", +, "5", "6", "-", "/", \*, "6", "/"]
Output: "12 * (11 + y)/(5 - 6) / 6"

Design a bit.ly like URL shortening service.

• service takes a URL and return a shortened version
• users can register on the service and previous shortened URLs

Round 2

You are given an 2-D matrix with nulls at some points and servers at others. Return a new matrix with the manhattan distance of the each point from the closest server.

Example:

Input:
[['x', 'x', 'x']
,['s', 'x', 's']
,['x', 'x', 'x']
,['x', 's', 'x']]

Output:
[['1', '2', '1']
,['0', '1', '0']
,['1', '1', '1']
,['1', '0', '1']]

Follow up:

How would you store the input matrix if the count of nulls was in billions, but the server were in thousands.

Round 3

Given a number N, return the minimum count of perfect squares that sum upto N.

Example

foo(6) returns 3 // 4 + 1 + 1
foo(41) returns 2 // 16 + 25

Follow up

Given a number N, return a array of perfect squares, with minimum count, that sum upto N.

Example

foo(6) returns [4, 1, 1]
foo(41) returns [16, 25]

Round 4

Design contraints

You are given a system with these properties:

• The system is a social media platform
• Each user gain points on the platform through various activities such as:
  • 5 points for signing up
  • 1 point for daily check-in
  • 3 points for referrals
• The system displays users' current points and their overall rank across all users on the websites' banner (always visible on the website)
• The banner should be kept up to date.
• There are about 200M total users.

Question

Design a point manager class with these APIs

- AddPoints(int pointsToIncrement, GUID userId)
- GetPoints(GUID userId)
- GetRank(GUID userId)

Follow-up

You may have assumed that pointsToIncrement is relatively very small compared to the range points.
You no longer have this assumption. The APIs runtime should be around logarthmic time.