[is0tope]

I used to work in a bank and wrote an interview script for my department (kdb+/q market data). It mainly consisted of having the candidate sit in front of an interpreter, and walk them through a script which looked something like:

1. Load a data file here

2. Tell me some facts about the data

3. Here's another dataset, can we use both to figure something out.

4. One of the executions for this order is missing, how can we find which one.

5. Here is a data feed, can you write a process to ingest the data and calculate something in real time.

I by far preferred this system to the alternative which was to ask trivia questions and see if the candidate memorised the docs. There is of course some value in asking basics, or to elaborate etc. But on the spot algo questions are usually only useful in filtering people who either like leetcode problems, or have grinded them for the last 6 months.