* leetcode, which has never once been relevant to my actual job in 20 years.
* during the data science uber alles days they'd ask me to regurgitate all sorts of specialized DS stuff that wasnt relevant before throwing me into a project with filthy pipelines and where picking a model took all of about 20 minutes.
* I remember the days when nosql and "scaling" was all the rage and being asked all sorts of complex questions about partitioning and dealing with high throughput while the reality on the ground was that the entire company's data fitted easily on to one server.
* More recently i was asked about the finer details of fine tuning llms for a job where fine tuning was clearly unnecessary.
I could go on.
It's been a fairly reliable constant throughout my career that hiring tasks and questions have more often been driven by fashion and crowd following than the skills actually required to get the job done and if you refuse to play the game at all you end up disqualifying yourself from more than half the market.
* during the data science uber alles days they'd ask me to regurgitate all sorts of specialized DS stuff that wasnt relevant before throwing me into a project with filthy pipelines and where picking a model took all of about 20 minutes.
* I remember the days when nosql and "scaling" was all the rage and being asked all sorts of complex questions about partitioning and dealing with high throughput while the reality on the ground was that the entire company's data fitted easily on to one server.
* More recently i was asked about the finer details of fine tuning llms for a job where fine tuning was clearly unnecessary.
I could go on.
It's been a fairly reliable constant throughout my career that hiring tasks and questions have more often been driven by fashion and crowd following than the skills actually required to get the job done and if you refuse to play the game at all you end up disqualifying yourself from more than half the market.