We all want "inclusive hiring culture", "exceptional engineering talent" and a frictionless hiring process, but IMO you can't have all three. Like them or not, leetcode* interviews actually give a chance to people who can't do a home assignment, or come from a background that didn't let them have a bunch of code on Github. In that sense, it's the more fair way to test people's aptitude, and will find exceptional talent from all kinds of different backgrounds.
If you still want "exceptional talent", but not algorithmic interviews, then you end up biasing towards white guys who have a ton of projects to show you.
Actually, I think this should be verifiable. Select some companies that we think have exceptionally high bar (you could use compensation as a proxy, acknowledging it's imperfect). Then classify them based on whether they do "leetcode" interviews or not, and check their diversity reports. My bet would be that the "leetcode" companies do significantly better.
* Caveat is that companies people think do "leetcode" actually usually ban questions that appear on leetcode.
> Like them or not, "leetcode interviews" actually give a chance to people who can't do a home assignment
Not really—if I don't have time to do an hour-long take-home assignment, what makes you think I have time to practice leetcode-style questions? The take-home assignment is usually testing skills that I actually use in my job on a regular basis, so I don't need extra preparation, I just need a block of time to sit down and do it.
I agree that expecting people to be able to show side projects is a mistake, though I'm not sure why you think that the bias there would be racial—I would imagine it would be much more a filter that excludes people with families and/or non-computer hobbies.
I guess my assumption is that measuring cognitive ability is better than specific skills. The algo interviews, when done well, do that better than interviews focused on experience.
There are of course a lot of caveats - experienced candidates should be treated differently from entry level. It’s also true that, e.g. Google has a weird fetish about dynamic programming questions and other problems that people are unlikely to figure out without having taken a class in them.
On the other hand, take home assignments take up time and room you might not have. It’s easier to do those when you’re a man in a developed country than, e.g. a single mother in Alabama.
So its basic, I think algo interviews are a good way to hire junior level engineers.
If you still want "exceptional talent", but not algorithmic interviews, then you end up biasing towards white guys who have a ton of projects to show you.
Actually, I think this should be verifiable. Select some companies that we think have exceptionally high bar (you could use compensation as a proxy, acknowledging it's imperfect). Then classify them based on whether they do "leetcode" interviews or not, and check their diversity reports. My bet would be that the "leetcode" companies do significantly better.
* Caveat is that companies people think do "leetcode" actually usually ban questions that appear on leetcode.