yoe: 3 as research sceintist in ML + NLP + IR

Applied with referral on careers website. Got an email from a recruiter for an initial phone screen with a team member.

(ML - machine learning, DS - Data Science, NLP - Natural Language Processing, IR - Information Retrieval)

First phone screen

Questions on resume and background, some basic ML questions. General questions and back and forth to gauage my interests in the role.

Second phone screen

• Shortly after the first interview, got an email from recrutier to scedule another phone screen with another team member.
• This one was pretty similar to the first phone screen but it was longer.
• A lot of questions on basic ML principles in different domains like basic ML, DL, NLP, computer vision, optimizers, bias vs variance, etc.

Moved to onsite after this, a loop of 3 interviews. Hiring manager, team member, skip manager.

Onsite loop

First Round Hiring Manager

• Background questions to see team fit.
• Discussion about ML basics like how are k-means and knn similar and different, so on
• Guesstimate based question like how many walmarts in US.
• Cultural fit based questions. Essentially collaboration is the key in workplace environment.
• Finally there was some to kill and was asked to code a math function which can be solved using binary search or Newton's method.
  This round went amazing, easy converstion.

Second Round

• This round with a senior DS in the team started of with introductions and some basic questions related to previous projects which the interviewer also has related experience with.
• Two questions from leetcode
  • First one was easy stream based question. Gave optimal O(1) time and O(stream limit) space solution.
  • Second one was cheesy. It is a medium question on leetcode which many NLP people use regularly. It is based on words and their frequencies. Gave optimal heap based O(nlogk) solution. (if you did not guess what this question is, I recommend doing all heap + word based medium problems on LC asap)
• This round went extremely well.

Final Round

• This was with skip manager/ director of the group.
• Background and culture fit questions.
• Questions to guage communication skill wrt stakeholders. For eg, explain your work to 90 year old.
• Design question - implement a data structure for optimal storing and retrieval.
  • Asked some clarifying questions about the data and eventually decided to implement hashtable with chaining
• This round was weird but overall it went well.

A week later, recruiter emailed me saying that the team has decided not to move forward with the application.

Prep:
LC: 750+ (~100 hard, rest are mostly medium), I do LC almost every day.
ML: My job requires paper reading so it is hard to stay out of touch. I also read blog posts, books regularly so that helps.

My outtake:
I don't know, there was no feedback given so it kind of sucks. The interview experience was smooth, somewhat problematic with scheduling but that's expected with the pandamic situation.
I probably will not apply to Microsoft again, this is because it feels like getting the job and doing well in interviews are widely different things. I don't know how much time I can keep wasting on interviews like these where the performance is not terrible but still get rejected without any feedback. How can one improve?

good luck to others.