Meta ML Engineer Interview – E5 (2024)

An anonymous candidate shared their interview experience for the E5 Machine Learning Engineer position at Meta (Menlo Park). The interview process spanned over multiple rounds across two days, involving both algorithmic coding and machine learning system design interviews.

Initial Contact

The recruiter initially reached out in June 2024, with formal engagement starting in January 2024. A phone screen was scheduled for early February, followed by a full onsite loop in April 2024, planned across two separate days to allow for optimal preparation.

Phone Screen

The candidate solved two standard LeetCode problems:

• Merge Intervals

• Validate Binary Search Tree
  
  

Both were completed just in time. The performance was strong enough to advance to the onsite rounds.

Onsite Interview

Day 1

Coding Interview 1:

• Question 1: Maximum subtree sum in a binary tree.
  The task involved evaluating each node as a subtree root and returning the maximum subtree sum.

• Question 2: Variant of Word Break
  The input string had to be segmented into a valid sentence using dictionary words, producing outputs like "cat sand dogs" or "cats and dogs".
  
  

Coding Interview 2:

• Top K Frequent Elements

• Course Schedule II
  
  

The candidate used efficient techniques such as heaps and topological sorting to solve the problems.

Day 2

Behavioral + Coding Interview:

• Behavioral questions focused on workplace collaboration, ownership, and conflict resolution.

• Coding Question: Find the length of the longest subarray whose sum equals a target value.
  A HashMap-based approach was used to solve this in O(n) time.
  
  

System Design (Travel Recommendation App):

• Design a machine learning system to recommend travel destinations during a user’s trip.

• Covered components:

  • Data sources and pipelines

  • Feature engineering

  • Offline and online training setup

  • Evaluation strategy

  • Feedback loops

  • Online A/B testing

  • Scalability and latency trade-offs

The candidate found this round challenging due to limited experience with mobile/backend ML system design.

ML System Design (Facebook Marketplace):

• Tasked with designing an E2E classification pipeline for Marketplace posts.

• Inputs included both image and textual descriptions.

• Key discussion points:

  • Feature fusion (multi-modal input)

  • Labeling strategies

  • Training and inference pipelines

  • Monitoring, retraining, and experiment tracking

  • Theoretical depth and real-world application

The candidate considered this their strongest round.

Calibration Round:

• Behavioral questions similar to earlier rounds

• Coding Question: Insert a node at the p-th position in a singly linked list.
  Required handling edge cases and pointer logic accurately.
  
  

Conclusion

Overall, the candidate had a positive experience. Despite limited preparation for system design and ML infra, the opportunity helped them calibrate expectations for high-level ML roles at top-tier companies like Meta.

Interestingly, they received a promotion at their current company shortly after the phone screen, which reduced performance pressure and allowed for a more relaxed approach during the interview loop.