Meta Interview Questions — Page 202

Meta·Machine Learning Engineer·Technical Phone Screen

Jun 2026

Meta ML Engineer interview with a coding round that leaned on the classics. Nothing too exotic, but the in-place constraint is the kind of thing that trips you up if you haven't thought about it recently.

You have two sorted arrays, nums1 and nums2, with nums1 having extra space at the end to hold all elements. Merge nums2 into nums1 in-place so the result is sorted, without allocating a new array.

“The trick is starting from the back, not the front.”

View Post

Meta·Software Engineer·Onsite - System Design / Architecture

Jun 2026

Meta system design round focused on building a LeetCode-style coding judge platform, which sounds scoped until you realize they want the full thing: sandboxed grading, user profiles, contest infrastructure, and a real-time leaderboard that doesn't fall apart under load. Solid question, took me a while to find my footing.

Design a coding judge platform like LeetCode, including problem catalog, code submission with sandboxed grading, user profiles, and contest support with real-time leaderboards.

“I started with the problem catalog and submission flow, which felt safe, but the interviewer kept nudging toward the contest side and I realized I'd been avoiding it because leaderboards under write pressure are genuinely tricky.”

View Post

Meta·Product Manager·Onsite - Product Sense / Strategy

Jun 2026

Meta PM interview, product sense round. Just one question but it was a cross-company one which felt a bit odd to get at Meta specifically.

How would you define and set success metrics for Google Maps?

“Weird to get a Google product question in a Meta loop but here we are.”

View Post

Meta·Software Engineer·Technical Phone Screen

Jun 2026

Meta SWE interview focused on extending a banking system with scheduled payment functionality. Pretty deep on the implementation details and edge cases, more than I expected for a coding round.

Design and implement a scheduled payment system for a banking application, including a function to schedule a future payment with a delay offset and a function to cancel a scheduled payment, with appropriate return values for each.

“The scheduling part was fine, global counter for payment IDs is straightforward.”

View Post

Meta·Data Scientist·Technical Phone Screen

Jun 2026

Meta DS interview with a SQL and metrics design question around marketplace shop visibility. Pretty applied, less about syntax and more about whether you can actually think through a business problem with data.

Given a shops table and a page_views table, write SQL to compute daily page views per shop. Then design and calculate a metric that shows whether shops created in the last 30 days get more visibility than older shops.

“The SQL part was fine, just a join and a group by date and shop_id.”

View Post

Meta·Software Engineer·Technical Phone Screen

Jun 2026

Meta SWE coding round, one tree problem. Pretty straightforward session, nothing too wild.

Given a binary tree, return the values of the nodes visible from the right side, ordered from top to bottom.

“Classic level-order traversal problem.”

View Post

Meta·Data Scientist·Technical Phone Screen

Jun 2026

SQL-heavy data science screen at Meta focused entirely on marketplace metrics. One long question that spiraled into about five sub-questions once you thought you were done. The kind of interview where you realize halfway through that knowing SQL isn't the same as knowing how to think about SQL at scale.

You own a 'shop visibility' KPI for a marketplace. Define a precise metric for it and write SQL to compute it over a 7-day window. You need two definitions: (A) visibility_rate as unique US users who saw at least one active product from a shop divided by total active US users, and (B) impression_share as impressions of a shop's active products divided by total impressions of all active products. Exclude deactivated products, deduplicate multiple impressions of the same product by the same user within a session, and restrict active users to those with at least one session in the window. Return shop_id, visibility_rate, impression_share, unique_viewers, active_users, window_start, and window_end.
How would you parameterize the same shop visibility query per surface, for example separating feed impressions from shoppage impressions?
How would you handle late-arriving impressions and bot traffic in this pipeline?
What indexes or table clustering strategies would you use to make this query efficient at over a billion rows?
How would you adapt this query for incremental daily jobs and backfills?

“The metric definition part felt fine at first.”

View Post

Meta·Software Engineer·Technical Phone Screen

Jun 2026

Coding round for an SRE role at Meta, two problems back to back. Both were pretty standard implementation stuff, nothing that required any fancy algorithmic knowledge, just clean logic and edge case awareness.

Given a word and an abbreviation string, determine whether the abbreviation is valid. Letters must match directly, digits indicate how many characters to skip, no leading zeros allowed, and the abbreviation must consume the entire word.
Given a string containing only bracket characters, return whether the brackets are balanced and closed in the correct order.

“The leading zeros edge case is the one that'll get you if you're moving fast.”

View Post

Meta·Machine Learning Engineer·Technical Phone Screen

Jun 2026

Meta ML Engineer interview with a pretty gnarly algorithms problem focused on k-th smallest pair sum. No behavioral fluff, just straight into the deep end on heap-based approaches and complexity proofs.

Given two sorted non-decreasing arrays of non-negative integers and an integer k, return the k-th smallest value among all possible pair sums (one element from each array), without enumerating all pairs. Target O(k log min(n, m)) time and O(min(n, m)) space using a heap. Also discuss duplicate sums, edge cases like k=1 and k=n*m, and prove your approach is correct.

“This one hurt a little.”

View Post

Meta·Data Scientist·Technical Phone Screen

Jun 2026

Tough technical screen for a DS role at Meta. The whole thing was basically one long Bayesian stats problem with a bunch of sub-parts, and they pushed hard on the practical implications, not just the math.

You have a binary KPI with Beta(1,1) priors and 10,000 users per arm. One arm gets 515 conversions, the other 500. Compute the posterior for each arm and the probability that lift is greater than zero.
How would you define a decision rule using expected loss when the costs of false positives and false negatives are asymmetric?
Compare this Bayesian decision approach to a frequentist test run under optional stopping. What are the operational differences, and how would you communicate posterior probabilities versus p-values to stakeholders?
How would you set or elicit priors in a way that avoids over-optimism in your experiments?

“The Beta(1,1) prior part is easy since it's just uniform, so the posteriors are Beta(516,9486) and Beta(501,9501).”

View Post