4 ms·

> It's better to have high standards to begin with to avoid needing to fire people. Which leads to the job interview system that doesn't satisfy anybody either

> It's better to have high standards to begin with to avoid needing to fire people. Which leads to the job interview system that doesn't satisfy anybody either.

adam_arthur 4y ago
The tech interview system employed by FAANG is an example of a weak correlation style interview. Actually the structure of the interview is mostly fine, but the criteria being judged is usually off. 1) Questions should be structured such that they're modeled in a real world context, and somewhat close to the nature of the type of problems your company/division solves. 2) For most companies, coding ability and quality is more important than CS theory strength. Success on a coding problem for these companies should be judged by pace of coding and quality of solution, rather than time complexity of the result. Run time complexity of an algorithm is almost 100% orthogonal to ability to write high quality code quickly, yet this is where 99% of the interview focus is for most companies. That being said, if you're hiring somebody to design a database storage system, sure, theory is more important in that context. But 99% of jobs are not that. Can't tell you how many people I've seen join FAANG that I've worked with who are actually quite poor performers in a real world context. It's very easy to grind leetcode and game the system as its structured. Its true too that at scale the correlation is probably good enough to end up with a decent workforce though. But also very easy to tweak judging criteria to be more highly correlated to real world success. I've hired close to 100 engineers, and its immediately obvious to me who will perform well by how they carry themselves in the interview. I pretty much don't even take into consideration whether they reach an optimal runtime solution. One of the best guys I hired couldn't even implement a tree traversal/DFS/BFS in the interview

> It's better to have high standards to begin with to avoid needing to fire people. Which leads to the job interview system that doesn't satisfy anybody either.

adam_arthur 4y ago

The tech interview system employed by FAANG is an example of a weak correlation style interview. Actually the structure of the interview is mostly fine, but the criteria being judged is usually off. 1) Questions should be structured such that they're modeled in a real world context, and somewhat close to the nature of the type of problems your company/division solves. 2) For most companies, coding ability and quality is more important than CS theory strength. Success on a coding problem for these companies should be judged by pace of coding and quality of solution, rather than time complexity of the result. Run time complexity of an algorithm is almost 100% orthogonal to ability to write high quality code quickly, yet this is where 99% of the interview focus is for most companies. That being said, if you're hiring somebody to design a database storage system, sure, theory is more important in that context. But 99% of jobs are not that. Can't tell you how many people I've seen join FAANG that I've worked with who are actually quite poor performers in a real world context. It's very easy to grind leetcode and game the system as its structured. Its true too that at scale the correlation is probably good enough to end up with a decent workforce though. But also very easy to tweak judging criteria to be more highly correlated to real world success. I've hired close to 100 engineers, and its immediately obvious to me who will perform well by how they carry themselves in the interview. I pretty much don't even take into consideration whether they reach an optimal runtime solution. One of the best guys I hired couldn't even implement a tree traversal/DFS/BFS in the interview