Re: Question: Regarding Creating An Tech-Candidate Assessment Application

A related method that my teams have used in the past is giving people a relevant design problem prior to the interview that will then be discussed in depth. The problems selected have an important feature: the most naive designs should be familiar to everyone and easily looked up but the most advanced designs and techniques are neither Google-able nor in literature. This allows the candidate to start from a comfortable foundation that can be built upon with guided questions about design limitations and possible alternatives that can’t be memorized ahead of time. This is works well for exposing how they think about design problems. (If they are familiar with the unpublished state-of-the-art then that is proof of domain expertise ipso facto.) 

As an example, for people hacking on database engines a problem is cache replacement algorithm design. Everyone knows about LRU designs, which are extremely naive, and a quick Google search will turn up ARC, CLOCK, etc algorithms that illustrate some of the limitations that need to be addressed, and there are some important “open” design problems (e.g. recency saturation, L1/L2 cache thrashing, et al) that have not-too-clever (if appropriately nudged) solutions that don’t exist in Google or literature. The sub-problems are sufficiently diverse in nature that it exposes much about the limitations of candidates' current skills, and no one is expected to reason themselves to a state-of-the-art design (though it would be impressive if they did).

There are problems like this everywhere. Computational geometry, storage, sharding, etc.


> On Mar 25, 2019, at 4:34 AM, Stephen D. Williams <[email protected]> wrote:
> 
> My most accurate technical interview method was to come up with a particularly good problem that would seem impossible to work out in an interview, then walk them through it.  At each step, let them make as much progress as they could with degrees of hints and knowledge.  That shows how quickly they can learn, adapt, incorporate and reason with new information.  This problem has a bit of physical reasoning, a series of linked data conversions, and a few algorithmic problems.  There are opportunities for creative insights, strategies, etc.  I think of the method as being Active Interviewing, Socratic Interviewing.  It gives me a great understanding of whether they are a true problem solver, true auto-didactic learner, their capacity for keeping constraints, rules, and facts in mind, and what techniques they know about.
> …
> Can anyone suggest problems that fit this?


_______________________________________________
FoRK mailing list
http://xent.com/mailman/listinfo/fork