XgBoost is tree based? — Interview tales

Few months ago, I interviewed with a consulting giant amongst big 4 (just for fun though). It was a technical manager on the other side. To be honest, that man’s arrogance was out of the world.

He started asking me about Chat completions API and claimed that embeddings are mandatory. I countered, saying the LLMs can create their own embeddings internally. But, he kept arguing about a few things he learnt by rote. He asked me about the formulae of some metrics, few of which I did not remember by rote.

Further down the line, I realised that this man knew nothing about AI/ML. He did not know that XgBoost is a tree-based algorithm and argued that max depth is not a hyperparameter.

In summary, the whole interview was about questions that tested my rote learning abilities, instead of applied skills. Anyway, their job, their choice.

But, it’s sad that these organizations, with such leadership mint billions of dollars in the name of AI and thus, investors lose confidence in AI as a whole.