Autocoder

Meta Interview Questions and Answers for SDE Roles

Last updated 2026-11-19

Meta interviews test problem solving, coding, debugging and collaboration, with a behavioural interview on past experience in the full loop. This page covers all of those with model answers, including how to approach its AI-assisted coding setup.

The questions are practice in the style of the topics Meta says it covers, not a record of past interviews.

ProcessRecruiter conversation, initial screen, full loop
Coding toolCoderPad for select roles, with built-in AI assistant
AssessedProblem solving, coding, debugging, collaboration
BehaviouralAt least one interview on past experience

How the rounds are described

Meta's hiring-process page lists a recruiter conversation, an initial screen and a full-loop interview, typically over two to three months. For select roles the technical interviews run in CoderPad with a built-in AI assistant that candidates are expected to use, and outside AI tools are not allowed. Older Meta preparation guidance describes about 35 minutes of coding on algorithms and data structures in the initial screen, and a full loop mixing coding, system or product design (about 45 minutes in that guidance) and at least one behavioural interview about past experience.

Meta says interviews assess problem-solving, coding, debugging and collaboration. Specific counts for India SDE roles are not published, so ask your recruiter for the schedule.

Related practice

Coding problems in the style of these rounds are in Meta coding round questions. For the full process see how to crack the Meta interview.

Hiring status

On 20 May 2026 Meta cut nearly 8,000 jobs and moved about 7,000 workers to AI work. In an internal memo reported by Reuters, Mark Zuckerberg wrote 'we do not expect other company-wide layoffs this year', which does not rule out team-level cuts. As background, in February 2025 Meta opened 41 engineering roles in Bengaluru (software, machine learning and data-centre chip design) while cutting staff elsewhere.

See Meta hiring for the sourced summary and Meta's own careers site, metacareers.com, for live roles.

Meta Interview Questions and Answers for SDE Roles: questions with worked answers

  1. 1

    Given a social graph, how do you find all friends within two hops of a user?

    medium

    Run a breadth-first search from the user and stop at depth two, using a visited set so each person is processed once. Time is O(V + E) over the explored region; with average friend count f this is roughly O(f squared). For large graphs, mention that you never load the whole graph and that fan-out explodes with depth. A common mistake is DFS, which does not respect hop distance, or forgetting to exclude the user and direct friends if the question asks only for friends of friends.

    python
    from collections import deque
    def within_two(graph, user):
        seen, q = {user}, deque([(user, 0)])
        while q:
            u, d = q.popleft()
            if d == 2: continue
            for v in graph.get(u, []):
                if v not in seen:
                    seen.add(v); q.append((v, d + 1))
        seen.discard(user)
        return seen
    java
    import java.util.*;
    class Solution {
        Set<Integer> withinTwo(Map<Integer, List<Integer>> g, int user) {
            Set<Integer> seen = new HashSet<>(List.of(user));
            Deque<int[]> q = new ArrayDeque<>();
            q.add(new int[]{user, 0});
            while (!q.isEmpty()) {
                int[] p = q.poll();
                if (p[1] == 2) continue;
                for (int v : g.getOrDefault(p[0], List.of()))
                    if (seen.add(v)) q.add(new int[]{v, p[1] + 1});
            }
            seen.remove(user);
            return seen;
        }
    }
  2. 2

    How would you find mutual friends of two users, and what does it cost?

    easy

    Put one friend list in a hash set and scan the other, giving O(a + b) time and O(min(a, b)) space if you put the smaller list in the set. If both lists are already sorted, use two pointers in O(a + b) with O(1) extra space. At scale you might precompute or cache mutual-friend counts for frequently viewed pairs. State both options and when each wins.

  3. 3

    What is the complexity of building a heap, and why is it O(n) rather than O(n log n)?

    medium

    Heapify bottom-up: sift down every non-leaf node starting from the last one. A node at height h costs O(h), and there are at most n / 2^(h+1) nodes at height h, so the total is the sum of h * n / 2^(h+1), which converges to O(n). Inserting n items one by one is O(n log n). Mention that heap push and pop are O(log n) and peek is O(1).

  4. 4

    Meta's coding interviews for select roles use a built-in AI assistant. How would you use it well?

    medium

    Meta's hiring-process page says technical interviews for select roles run in CoderPad with a built-in AI assistant that candidates are expected to use, and that outside AI tools are not allowed. Treat it as a pair programmer: state your plan first, ask it for boilerplate or to suggest edge cases, but read and verify everything, because interviewers assess your problem-solving, coding, debugging and collaboration. Check with your recruiter which rules apply to your role.

  5. 5

    Design a news feed at a high level.

    hard

    Clarify scope: posts from friends, ranking, scale. Two approaches: fan-out on write (push each new post into followers' feed caches, fast reads, costly for users with millions of followers) and fan-out on read (assemble the feed at request time, cheaper writes, slower reads). A hybrid pushes for normal users and pulls for celebrities. Add ranking as a separate service, pagination with cursors, and a cache layer. See system design interview.

  6. 6

    A service starts returning errors after a deploy. How do you debug it?

    medium

    Meta says its interviews assess debugging. Answer with a method: confirm the symptom and its blast radius, check what changed (deploy, config, dependency), compare error rates before and after, roll back first if users are affected, then read logs and traces to find the failing call, reproduce in a test, and add a regression test and monitoring. Show that you reduce user impact before hunting for root cause.

  7. 7

    When is memoisation better than bottom-up dynamic programming?

    medium

    Top-down memoisation only computes states that are actually reached, so it can be faster when many subproblems are unreachable, and it is closer to the natural recurrence. Bottom-up avoids recursion depth limits, has lower constant overhead and allows space optimisation by keeping only the previous row. For Fibonacci-style problems bottom-up with two variables uses O(1) space. Mention converting between them as a skill. See dynamic programming interview questions.

  8. 8

    Why is repeated string concatenation in a loop slow?

    easy

    Strings are immutable in Java and Python, so each concatenation can copy the whole string, making n appends cost O(n squared) in total. Use StringBuilder in Java or collect parts in a list and join once in Python, which is O(n). CPython sometimes optimises in-place concatenation, but you should not rely on it. This matters in problems that build output strings character by character.

  9. 9

    Tell me about the most impactful project you worked on.

    medium

    Meta's process includes at least one behavioural interview on past experience. Lead with the outcome and why it mattered (users affected, time saved, error rate), then your specific role, the hardest decision and how you measured success. Use numbers you can defend. If your impact was small in scale, be clear about what you learned and how you would scale it. Avoid listing technologies without saying what you owned.

  10. 10

    Tell me about a time you received difficult feedback.

    easy

    Choose feedback that was true and that you acted on. Describe what was said, your first reaction honestly, what you did to verify and fix it, and how your behaviour changed afterwards, ideally with a later example showing the change. Interviewers look for self-awareness and for you not becoming defensive. Use the structure in the behavioural interview guide.

  11. 11

    Tell me about a time you had to move fast and cut scope.

    medium

    Describe the deadline or opportunity, how you decided what to drop, who you consulted, what shipped and what happened after. Good answers show a clear rule for cutting, such as keeping the path users must have and deferring polish, plus how you managed the debt you created. State what you would do differently so the story shows judgement rather than luck.

  12. 12

    Describe a conflict with a teammate and how you resolved it.

    medium

    Explain the disagreement neutrally, what each side wanted, how you listened and checked your assumptions, and what you agreed. Focus on actions you took, such as a short one-to-one conversation or a written comparison of options, and the effect on the project. Avoid stories where the other person was simply wrong; interviewers read that as low collaboration.

Frequently asked questions

Does Meta allow AI in coding interviews?

Meta's hiring-process page says select technical interviews use CoderPad with a built-in AI assistant candidates are expected to use, and that outside AI tools are not allowed.

How long does the Meta hiring process take?

Meta says the process typically takes two to three months from the recruiter conversation to a decision. Timing varies.

What topics does Meta list for coding?

Algorithms and data structures, recursion, binary trees, graphs with BFS and DFS, hash tables, heaps and big-O analysis.

Where do I find Meta's current openings?

See Meta hiring and metacareers.com.

Sources

Related guides

More in Company Interview Questions

See all in this section