← Back to Directory

Meta interview notes, page 98 of 238

FAANG / Big Tech

Most of these Meta write-ups come from Software Engineers (1,194), then Data Scientists (489) and Product Managers (453). The most reported rounds are the technical phone screen (1,190), the product sense round (466), the system design round (255) and the behavioral / leadership round (205).

2377 interview notes · updated Jul 2026

Meta·Machine Learning Engineer·Technical Phone Screen

May 2026
Meta ML engineer interview with a grid-based coding problem. Not a lot of context in what I remember but the problem itself was meaty enough to keep me busy.
  • Given a 2D grid of 1s and 0s representing land and water, find the largest perimeter among all islands in the grid.

“Classic grid traversal but the perimeter part tripped me up a bit.”

View Post

Meta·Product Manager·Onsite - Product Sense / Strategy

May 2026
PM product sense question at Meta, just the one question but it's the kind that can spiral in a dozen directions if you're not careful.
  • You're the PM for Safeway. What product would you build to improve how the store serves its customers?

“I went straight to a loyalty/personalization angle and kind of talked myself in circles.”

View Post

Meta·Machine Learning Engineer·Onsite - System Design / Architecture

May 2026
Research-design round at Meta for an MLE role, focused entirely on building an LLM-based conversational assistant from scratch. Pretty grueling scope: base model selection, alignment pipeline, RAG and tooling, then serving and safety all in one session.
  • Walk through how you'd design a general-purpose LLM-based conversational assistant end to end, covering base model choice, alignment, grounding, and production serving.
  • What clarifying questions would you ask before committing to an architecture, and what are the key requirements you'd pin down first?
  • How do you turn a pretrained base model into a safe, instruction-following assistant? Describe the post-training pipeline and the data you'd need.
  • How do you keep the assistant grounded in fresh or private information it never saw during training?
  • How do you evaluate assistant quality, and how do you serve streaming responses at scale within a tight latency budget?
  • Users say the assistant is too cautious and refuses benign requests. How do you measure over-refusal and fix it without weakening real safety?
  • The model hallucinates citations even with retrieval enabled. How do you figure out whether the problem is in retrieval, reranking, or generation?
  • How would you let an enterprise customer ground answers in their private corpus while guaranteeing that data never enters your training set or leaks to other tenants?
  • How would you A/B test a new alignment recipe in production without exposing users to a safety regression?

“This is basically four separate design problems stapled together, and I felt the time pressure immediately.”

View Post

Meta·Product Manager·Onsite - Product Sense / Strategy

May 2026
Meta product sense interview focused on the Newsfeed. Pretty classic PM metrics question but the follow-up about trade-offs is where things get real.
  • What metrics would you use to measure Facebook Newsfeed usage, which one should be the primary metric, and what trade-offs come with that choice?

“This looks straightforward until you try to pick just one top metric and defend it.”

View Post

Meta·Product Manager·Recruiter / HR Screen

May 2026
Interviewed at Meta for a product role, got asked the classic favorite-product question. Pretty standard screen, nothing too wild.
  • What is your favorite Facebook product and why?

“I went with Marketplace because I had a real story around it, but in hindsight I wonder if that was too safe.”

View Post

Meta·Data Scientist·Technical Phone Screen

May 2026
SQL-heavy technical screen for a DS role at Meta. Two questions, both on the same schema, with the second one building on the first. Nothing conceptually wild but the filtering conditions and the multi-account definition tripped me up a bit.
  • Using only the accounts table, write a query that buckets users by how many accounts they own: exactly 2, exactly 3, or 4 or more. Return the bucket label and the count of users in each bucket.
  • Among users who have at least two accounts, what percentage have at least one unread notification across any of their accounts? Use a cutoff date of 2025-01-01 and treat a notification as unread if its read_at is NULL at that point.

“Straightforward once you see it as a two-step group-by.”

View Post

Meta·Machine Learning Engineer·Technical Phone Screen

May 2026
Meta ML Engineer interview with a graph problem that sounds straightforward until you realize they want a specific algorithm by name. One question, but it had layers.
  • Given a list of airline tickets as [departure, arrival] pairs, reconstruct the full itinerary starting from JFK using all tickets exactly once. If multiple valid itineraries exist, return the one that comes first in lexical order.

“I knew this was an Eulerian path problem pretty quickly, which felt good.”

View Post

Meta·Software Engineer·Technical Phone Screen

May 2026
Meta phone screen for a software engineer role, just one coding question the whole time. Pretty straightforward string manipulation problem but the edge cases kept me second-guessing myself.
  • Given a sentence as a string of space-separated words, convert it to 'Goat Latin' by applying a set of transformation rules: words starting with a vowel get 'ma' appended, words starting with a consonant have their first letter moved to the end before appending 'ma', and then each word gets a number of 'a' characters equal to its 1-based position in the sentence.

“I got the basic logic pretty fast but the indexing tripped me up.”

View Post

Meta·Product Manager·Onsite - Product Sense / Strategy

May 2026
Meta PM interview with a product design question about building an app for people new to a city who need to find doctors. Pretty open-ended, which I wasn't fully prepared for.
  • Design an app that helps people who have recently moved to a new city find doctors.

“I jumped straight into features and had to backtrack when they asked me who exactly I was designing for.”

View Post

Meta·Software Engineer·Onsite - Coding / Algorithms

May 2026
Meta SWE coding round focused entirely on a bitmask problem that kept escalating in scale. Four sub-tasks, each one requiring a fundamentally different approach, and the interviewer was pretty rigid about which optimization path they wanted.
  • There's a bug in the starter code for the unique-character subset problem. It either miscounts on empty input or handles duplicate-letter words wrong. Find and fix it before doing anything else.
  • For a small word list (around 12 words), implement a backtracking solution that finds the subset of words with all unique characters across the combined set. Return the actual subset, not just its size.
  • Scale the solution to handle 100-200 words. What pruning strategies make backtracking feasible at this size?
  • Now scale to tens of thousands of words. Backtracking won't cut it. What's your approach, and can you still reconstruct the actual optimal subset?

“The annoying thing is you want to jump straight to the algorithm and they make you slow down first.”

View Post