Meta ML Engineer interview typically runs 4 rounds: screening, technical coding, ML system design, behavioral. Timeline is usually several weeks and the process is highly predictable and coding-heavy.
$168K
Avg. Base Comp
$253K
Avg. Total Comp
4
Typical Rounds
3-6 weeks
Process Length
Our candidates report that Meta’s ML Engineer loop is less about proving you can “do ML” in the abstract and more about showing you can think and code with the speed of a strong software engineer. A recurring theme is the tight time pressure on short, LeetCode-style problems: trees, graphs, DFS, DP, sliding window, and binary search tree variants show up repeatedly, often with only a small twist. Multiple candidates noted that the biggest trap was spending too long searching for the perfect approach instead of getting to a workable solution quickly and then tightening it up.
What’s non-obvious is how much the company seems to value algorithmic cleanliness even when the role is ML-focused. We’ve seen candidates get sparse-matrix optimization questions, binary tree conversion tasks, and calculator-style implementations that reward precise mental execution because there’s no runtime safety net. That means Meta is watching for bug-free reasoning under pressure, not just familiarity with common patterns. One candidate even called out that the interviewer framed the questions as representative of the work, which lines up with the broader signal here: they want people who can ship reliable code, not just talk about models.
The ML-specific portion is where candidates often underestimate the bar. The system design prompts skew toward recommendation, ranking, search, and safety-style problems like unsafe content detection, and the strongest responses connect model choices to product tradeoffs and efficiency constraints. We’ve also seen the behavioral conversation stay comparatively light, so the real separator is usually whether you can handle the coding pace and then pivot into a crisp, practical ML design discussion without drifting into generic theory.
Synthetized from 2 candidates reports by our editorial team.
Had an interview recently?
Share your experience. Unlock the full guide.
Real interview reports from people who went through the Meta process.
The hardest part for me was realizing how much the Meta process is a time game. After the recruiter reached out, I had an initial screening and then a technical interview with two LeetCode-style problems in 45 minutes. Both were around easy to medium difficulty, but they could still eat up time if you spent too long brainstorming. In my case, one of the questions was a binary search tree problem, and another involved DFS. I also saw a string/parentheses-style question and an integer-to-string type problem in the same general bucket, so it felt very much like Meta-tagged coding questions with a small twist rather than anything exotic. The main advice I’d give is to get to a workable solution quickly, especially on the first problem, because leaving it blank would hurt you more than having a rough but correct approach on paper.
The onsite was pretty standard: two coding rounds, one ML system design round, and one behavioral. The coding rounds followed the same pattern as the screen, with multiple short problems rather than one long one. The system design round was the hardest part for me, and it was clearly ML-focused, with recommendation and ranking-style design being the kind of thing I’d expect people to prepare for. The behavioral round was smoother and more conversational. Overall the interviewers were mostly professional, though one coding interviewer came across as rude and unprofessional, which made that round feel worse than it probably should have. I didn’t get an offer, but the process itself was very predictable once I understood the format. If you’re preparing, I’d focus on Meta-tagged coding questions, practice moving fast on medium problems, and spend real time on ML system design rather than treating it like a generic design interview.
Prep tip from this candidate
Focus your coding prep on Meta-tagged BST, DFS, string manipulation, and integer conversion problems, and practice solving them quickly since the format packs multiple short problems into 45 minutes — getting a rough but correct solution fast matters more than perfection. For the onsite, treat the ML system design round as its own distinct discipline by studying recommendation and ranking system design specifically, not generic system design frameworks.
Share your own interview experience to unlock all reports, or subscribe for full access.
Sourced from candidate reports and verified by our team.
Topics based on recent interview experiences.
Featured question at Meta
Given two sorted lists, write a function to merge them into one sorted list.
| Question | |
|---|---|
| Scrambled Tickets | |
| Find Bigrams | |
| One Element Removed | |
| Search Ranking | |
| P-value to a Layman | |
| Detecting Firearm Sales | |
| Level Of Rain Water In 2D Terrain | |
| Nearest Common Ancestor | |
| Facebook Stories | |
| Random Forest Explanation | |
| Bank Fraud Model | |
| Recurring Character | |
| Impression Reach | |
| Type-ahead Search | |
| Lazy Raters | |
| Find Duplicate Numbers in a List | |
| Hurdles In Data Projects | |
| Binary Tree Conversion | |
| Booking Regression | |
| Fill None Values | |
| Friendship Timeline | |
| Duplicate Rows | |
| Good Grades and Favorite Colors | |
| Intersecting Lines | |
| Dice Worth Rolling | |
| Spam Classifier | |
| Z and t-Tests | |
| Fake Algorithm Reviews | |
| Using R Squared |
Synthesized from candidate reports. Individual experiences may vary.
The process starts with an initial recruiter screening after the recruiter reaches out. This is a brief first contact to confirm interest and move candidates into technical interviews; in the experiences shared, there was little warm-up before the coding rounds began.
Candidates then complete a coding-focused phone screen with LeetCode-style problems. Interviewees reported one or two questions in this round, including easy-to-medium problems and sometimes a harder one, with topics such as binary trees, DFS, binary search trees, strings/parentheses, calculator-style problems, and sparse matrix optimization.
The onsite is described as a standard loop with multiple interviews. It typically includes two coding rounds, one ML system design round, and one behavioral round; the coding rounds often mirror the phone screen with multiple short problems, while the ML system design round focuses on recommendation and ranking-style design.
After the onsite loop, Meta makes a final hiring decision. Based on the experiences shared, candidates may wait some time after the technical rounds before hearing back, and the outcome is communicated as an offer or no offer.