Candidate Interview Evaluation: Fill & Download for Free

I’ve sat on a Google hiring committee for a while and have seen a fair number of candidates trying for the second time. Second-try candidates have the same hiring bar as everyone else. The results of the first interview slate are available to the committee but typically don’t come up. I often don’t even look at them to prevent biasing myself, since if the candidate would be hired off their most recent slate that’s all that matters.The only times I can remember discussing the first slate are in situations where the second slate had some hole in the questions that were asked making it difficult to evaluate some specific trait of the candidate. In those cases, I’ve looked back to the first slate to see if there was evidence there.

I think about diversity in the hiring process in a few ways:Are my job descriptions truly accurate, or do I list “required” skills and experience I don't actually need? For various reasons, men tend to be more willing to apply to jobs for which they don't actually meet the requirements on paper. It's also too easy to give off a certain cultural vibe like “bro” or “nerd supremacist” that will turn good candidates away.Am I getting a diverse set of resumes? Asking my recruiters to get more diverse resumes forces them to look in different places than usual, and that leads to finding unnoticed superstars. You can't find a diamond in the rough if you never look.Are interviews being conducted in a fair way? This is good for all candidates, but especially for those who are different from existing employees in a noticeable way.Are interviews evaluating the candidate's attitude towards inclusion? Again, inclusion is good for everyone, but especially affects those who are far from the group average. If I hire an uninclusive candidate, that person will tend to drive away people who are in the minority in some salient dimension, because those people already feel “different”.Are we effectively selling the job to successful candidates with other offers who might have concerns around bias and inclusion?I view diversity in hiring as a competitive advantage. If I am more effective at finding good resumes, I'll interview more good candidates. If I'm more fair in evaluating candidates, I'll make fewer bad hires and more good ones. And of course when I have a competitive offer, I want to win!It has worked really well for me. I've hired several people who are rock stars but who we never would have found if my team hadn't pushed for more diverse resumes.

You probably won’t be asked to implement a min-max or stable matching or “find the Hamiltonian cycle” algorithm. Questions like these test prior knowledge whereas the actual interview questions are more targeted to skill evaluation. The interviews evaluate your ability to create and analyze algorithms, not your knowledge of existing algorithms.So, for instance, there’s the classic egg-drop problem. You could use a linear search - start with floor 1 and keep going. That could take up to 99 trials, though. Another approach would be to use dynamic programming. Suppose you know the minimum number of trials for all buildings of height k-1. You could then calculate the answer for height k by considering the combined number of trials from height a and k-a. The smallest sum is the minimum number of trials. This algorithm is O(n^2). There’s also a way to come up with the answer through pure analysis. No code required.Another example involves random number generation. Given a function rand7() that produces uniformly random numbers from 0–6 implement rand5() that produces uniformly random numbers from 0–4. It’s actually not possible to do this and bound the running time simultaneously. A good candidate might suggest an approximate solution (e.g. evaluate (rand7() + 7 * rand7() % 5) which produces a non-uniform distribution) and extend it to an arbitrary-precision solution. They might also come up with a solution that calls rand7() repeatedly until it gets a value from 0–4. They should be able to analyze such a solution.In short a good candidate will be able to analyze the running time of their solution and provide incrementally better algorithms.