LinkedIn Senior AI engineer, Bay Area (Rejected)

I have 6.5 years of experiewnce in ML

First technical round went well. One coding question and some design questions. Dont remember the exact question but it was something like a list of numbers and i had to find a set based on prefix sum idea.

Virtual round (5 rounds)

Coding round

Question was a list of tuple is given example

[(abc,START,10), (def,START,20), (def,END,30), (abc,END,40)]  

<func_name, Start/end, time>

Find the inlcusve time and exclusive time for a function like abc
inclusive time is abc start time - abc end time
exclusive time is (40-10) - (30-20) for this example.

I solved it pretty good since if you watch it closly then it converts to alternate plus and minus for the exclusive time calucation. Was able to solve the variants also and answered all questions.
Variants were if abc is repeated and if the start and end of functions are random.

Statistics

I messed this one up (hear me out)
The question was

Assume there is a list of numbers which is basically points on the x axis  
You have to find the point with the minimum distance travel from all the points  

I started on a bad note (i dont think its that bad but i was not able to convinve the interviewer that it was not a bad approach). I said, assume its a one dimenestinal k-means problem and i will choose any random number as the centroid and will optimize from there by calucating the mean and moving the centroids till it stops moving. Interviewer started asking why i am using k-means here. I said, just think with me, if it will not work then we will try a different approach but he was not having it.

Basically the answer is if the points are even, then median is the min point, if its odd then any point between the middle two are min point. I used almost 45 mins to come to this solution and after iterations.

This is what costs me my interview.

Data Mining Product Design

This went well, the interviewer asked me about ways to improve the recommendation of existing linkedin given some new features and limitations and asked me to focus more on the data and model. It was a good 1 hr talk and i explained everythign from data gathering, to cleaning, to model selection to training, validation (offline, online)

HM call

Same old discussion about the projects , how i did, what i did, in details from models to data to every decision that i took

Data Mining

The interviewer was a Staff scientist (same as statistics call) and he asked me EVERYTHING (linear, logistics, tree, NN, RN, LSTM, LLM, all types of offiline validations for different types of models like classification, ranking, forecasting, NLP, NN, regularization methods) I was able to answer all the questions and he mentioned that I answered good and he was finished up with all the questions 15 mins before the end of the call and we had a very good discussion about his projects and he asekd me suggestions on how to add a new vertical in linkedin search, i asnwered and he was convinced with it (he said that).

I got a generic rejection email next day without any feedback.