RepostedLinkedIn logo

LinkedIn · Phone screen → Onsite

Offer

LinkedIn Senior MLE Interview Questions: 1 Experience

Machine learning engineer · Full-timeMaster's degreeReported Nov 19, 2026
Shared byjanebeingjane

A 2026 LinkedIn senior MLE loop that ended in an offer: LC 21 and ML basics by phone, then five onsite rounds, the hardest a squared-distance DP.

At a glance

  • Process: recruiter contact (May and late June 2026), a phone screen in August, five 1-hour onsite rounds in early September, and an offer in mid-September (report 01)
  • Asked recently: LC 21; a squared-distance version of LC 1478 with a mean-versus-median proof; a recommendation system design; ML knowledge questions (report 01)
  • Difficulty: the candidate rated the loop hard, but only the onsite coding round with the mailbox variant felt hard; the other rounds went fine
  • Practise first: Allocate Mailboxes with Squared Distance, Merge Two Sorted Lists, Design a Personalized Recommendation System

Coverage: One experienced machine learning engineer changing jobs, with a Master's degree, who was brought in by an agency recruiter. The job title was Software Engineer - AI, which the candidate describes as a senior MLE role. The video interviews ran from August to early September 2026, and the report went up on November 19, 2026. It gives no location.

What came up

In this one loop, coding came up in the phone screen and in two onsite rounds, and ML knowledge in the phone screen and one onsite round.

Topic Reports Last seen
Coding: DP and linked lists 1 of 1 Sep 2026
ML knowledge questions 1 of 1 Sep 2026
ML system design: recommendations 1 of 1 Sep 2026
Behavioral and resume projects 1 of 1 Sep 2026

Questions we found

Algorithms / Data Structures

Question Reports Last seen
Allocate mailboxes with squared distance (LC 1478 variant) 1 Sep 2026
Merge two sorted lists (LC 21) 1 Aug 2026

System Design

Question Reports Last seen
Recommendation system 1 Sep 2026

ML knowledge

Question Reports Last seen
Neural networks, in depth 1 Sep 2026
Decision trees 1 Sep 2026
Overfitting and underfitting 1 Aug 2026
Dropout 1 Aug 2026
Regularization 1 Aug 2026

1 detailed candidate timeline, loops, and screens

01: Senior MLE (Software Engineer - AI), full loop to an offer

Interview period: August to September 2026 · Report date: November 19, 2026 · Role: Software Engineer - AI (machine learning engineer), senior · Degree listed: Master's · Channel: Recruiting agency · Format: Video · Candidate's rating: Positive experience, hard · Result: Offer

Recruiter contact (May to June 2026)

  1. A recruiter first reached out in May 2026. After a short chat, the recruiter said a position would open in July and promised to get back in touch then.
  2. At the end of June, the recruiter came back and offered to set up interviews.

Phone screen (August 2026)

  1. Coding: LC 21, merging two sorted linked lists into one sorted list. FastPrep's practice version is Merge Two Sorted Lists.
  2. ML basics: a deep dive into neural networks, then a sweep through most of the ideas in an intro machine learning course (the candidate compared it to Andrew Ng's ML course). Overfitting and underfitting, dropout and regularization were among them.

Onsite (early September 2026): five rounds of 1 hour each, on one day or split over two

  1. AI coding round. LinkedIn had not yet moved this round to coding with AI tools, so it was a coding question built on an AI idea. The task was LC 1478 (Allocate Mailboxes) made harder. Each house's cost is its squared distance to the nearest mailbox (L2) instead of the plain distance (L1), and the goal is the smallest mean squared distance. The candidate said a plain DP is not enough here; the best solution adds running totals (prefix sums) of the positions and of their squares. A second hard part was proving to the interviewer why, with squared distance, a group's mailbox belongs at the mean of its houses and not the median. The follow-up asked for something simpler if the answer didn't have to be optimal, and the candidate suggested k-means clustering. The candidate says they could only pass because they had read about this problem in an earlier candidate's report. FastPrep's practice version is Allocate Mailboxes with Squared Distance. It asks for the total squared distance instead of the mean; the best placement is the same.
  2. Coding round. An easy-to-medium question the interviewer had written. The candidate no longer remembers it.
  3. System design round. A recommendation system: retrieval of candidate items, ranking, features, training samples, and offline and online metrics. The interviewer moved from topic to topic freely and also asked about the resume, since the candidate comes from a recommendations background. FastPrep's practice version is Design a Personalized Recommendation System.
  4. Behavioral and hiring manager round. About half behavioral questions and half questions about projects on the resume. The report names no single question.
  5. ML knowledge round. Decision trees and neural networks. Then the interviewer shared a screen showing LinkedIn's own list of ML questions and went through about five of them one by one. To the candidate, it felt like a university exam.

Result

  1. Only the AI coding round was hard; the others went fine.
  2. The offer came in mid-September 2026.

Common questions

What is the LinkedIn MLE AI coding round like?

In one 2026 senior loop, it was a coding question on an AI idea with no AI tools: a squared-distance version of LC 1478, a proof that the mean beats the median, and a k-means follow-up (report 01).

How many onsite rounds are there for a LinkedIn senior MLE?

One candidate had five 1-hour rounds: AI coding, coding, ML system design, behavioral with the hiring manager, and ML knowledge, on one day or two (report 01).

How long does the LinkedIn MLE hiring process take?

For one candidate it took about four months from the first recruiter message in May 2026 to the offer in mid-September; the interviews themselves fell in August and early September (report 01).

What to practise

Keep practicing

Sources: Anywhere useful interview info can be found.

0 total reactions. Your reactions: none.
0 comments

Comments

0

No comments yet — be the first to share what you know.