LinkedIn Machine Learning Engineer Interview Experience — Weird Phone Screen with BQs, Array Sum Coding and InMail System Design

Sharing a pretty weird LinkedIn phone screen.

A recruiter reached out in early September and set up a 20-minute call. We briefly talked about my background, the target role and salary, and then the phone screen got scheduled.

Phone screen: 1 hour.

It started with about 5 minutes of intro. I got interrupted before I even finished my self-introduction. Overall it felt like the interviewer wasn't very interested in my background, and he quickly said "move on to the next part".

Then 3 behavioral questions:

• Tell me about a time you launched a project with a tight deadline
• Tell me about a time you disagreed with your manager
• Tell me about a time you guided a junior scientist through a piece of work

Right after the BQs it went straight into coding.

Coding: Multi-dimensional Array Traversal

You are given a MultiDimArray object m that represents an array with an arbitrary number of dimensions. It provides two interfaces:

• getDims(): returns the size of each dimension. For example, for a 2 x 2 x 3 array it returns [2, 2, 3]
• get(indices): returns the element at the given indices. For example, get([0,1,2]) returns the value at that position

Implement:

def arraySum(m: MultiDimArray) -> int:

Right after coding there was a system design question.

System Design: Personalized LinkedIn InMail Generation

The background is LinkedIn Recruiter. Recruiters need to find candidates who match a hiring requirement among 1 billion+ LinkedIn users, and they may need to send InMail to hundreds of candidates every day. The problems:

• Writing a personalized message by hand for every person takes a huge amount of time
• A generic template is efficient but has no personalization
• Not enough personalization can hurt the candidate's willingness to reply and the overall hiring outcome

So the task is to design a system that helps recruiters efficiently generate personalized InMail based on the candidate's information and the job requirements.

In the reverse-questions part I asked the interviewer about the technical direction of their team, and he told me directly that they hadn't designed it yet and were still at the opening discussion stage. It felt like he didn't really want to tell me more.

What stood out is that the whole time there was almost no discussion of my actual research / ML background, and I didn't even get to finish the intro before we jumped into the BQs. Time was also extremely tight, so there was no room to think things through or explain them carefully. I don't know if this is just LinkedIn's style. Most likely I'm done for.