This is definitely different compared to how a real interview will be conducted. The candidate knows many details about the system and tends to dive into many details at first iteration. This is quite uncommon in real world. Usually, you will lay down the big components and gradually dive into more detailed level information.

Would help a lot if there wasn't any of the 'subscribe bells ringing', some of what the interviewee says is hard to understand, and the bell makes it even harder to understand.

Okay, first off, from my experience, this is not how a real-life interview is conducted. This video looks more like a Paper Review/Discussion than an Interview. The Interviewer is staying silent listening to the Interviewee going on and on, and he apparently has no feedback to give or steer the interview towards solve the problem when the candidate is spending a lot of time on the theory.

This has to be the most helpful and impressive data science video I've ever seen.

Why is everybody calling him Sachin?

Why is the drawing and explanation on the drawing pad so bad? It loses the quality of what he is saying. If you don't look at the drawing and listen, he is saying gold...

Requestion for session 2 with the Math part behind the scene

probably the best ML system interview I have seen till date :) 👍

It does not looks like how we do real life design interview, we need to spends at 5-7 mins for questions to sort the HLD and features needed.

If it wasn't for the HRNN approach, what is traditionally expected from the algorithm side?

First of all this is not at all an interview setting. Interviewer is the one who should be guiding the interview, this is simply one way talk. Interviewee just keeps on talking about 'n' different things at a time making it a presentation rather than an interview. Appreciate the wealth of knowledge but in an interview setting you should always go in a proper step by step process instead of jumping across topics. I think here breadth is covered but depth is not, generally in ML design you are required to talk about model in a proper manner, just mentioning a model from paper does not suffice.

Why is MRR metric better for mobile use cases, whereas NDCG and MAP better for desktop use cases?

system design, isnt it qps, service, storage, scale

Wow, this is fantastic. Thank you.

Excellent

Is this about how an ML design interview SHOULD NOT be conducted? The interviewee talks like a professor with his snide "right?" comments every now and then. Could also help if you took 2 extra seconds and chose a proper drawing software rather than make a gibberish puke of scribbles all over. And possibly slow down your speed to go over system components rather than stuff like HRNN etc.

I would definitely give this candidate no hire if he is interviewing for a developer’s position, even for a pm position, I don’t think this guy can articulate the logic behind the architecture to his team😂, no hire either 😂

Writing so bad; I want to puke