Meta Senior Software Engineer Interview Questions

1.Basic questions on background

2.Tell me about a time you had to dramatically cut scope to hit a launch date. What did you cut, who did you have to align, and do you stand by that call in hindsight?

3.Why do you want to join meta

4.Describe a time you pushed your team to take a technical approach that most people thought was too risky or too ambitious. How did you make the case, and what happened?

5.Walk me through how you decided what to work on in the last six months. Were there things on the roadmap you pushed back on or deprioritized? What was your reasoning?

6.Tell me about a time when you identified and drove a significant technical improvement that wasn't on the roadmap.

7.Describe a time when you had to make a decision that would affect multiple teams. How did you approach it?

8.Tell me about a time you had to deliver feedback to a peer or senior colleague that you knew they wouldn't want to hear. What was the feedback, how did you deliver it, and what happened to the relationship afterward?

9.Has there been a feature or system you built that you later realized had an unintended negative impact on users or communities? How did you discover it, and what did you do?

10.You're building a feature on Threads where users can follow topics, and you need to fan out new posts to all followers in near real-time. At Instagram scale, some topics have 50 million followers. How do you think about the fan-out architecture?

11.Your React Native component on Instagram's profile screen is re-rendering on every keystroke in an unrelated search bar, causing visible jank on low-end Android devices. How do you diagnose and fix this?

12.You're on-call and TAO starts throwing elevated error rates on reads for the social graph. P99 latency on News Feed load doubles within five minutes. Walk me through your first 30 minutes of investigation.

13.Instagram's feed ranking pipeline needs to score 500 candidate posts per user request in under 50ms. Your team owns the lightweight ranking model that runs first. How do you think about the latency-quality tradeoff in that stage?

14.Meta's Memcache layer is the primary read cache for News Feed. You're designing a new feature where users can mute posts from specific advertisers — this preference needs to be reflected in every feed load within seconds of being set. How do you handle cache invalidation here without torching your hit rate?

15.You're the tech lead for a team that owns a shared Relay data-fetching layer used by eight product teams at Meta. Two teams have started bypassing it and making direct REST calls because they say Relay is too slow to iterate on. How do you respond?

16.WhatsApp needs to support message delivery receipts at 100 billion messages per day. The current design uses a polling model and it's not scaling. How would you redesign the delivery receipt system?

17.Your team's GraphQL endpoint for fetching a user's friend list is being called with N+1 query patterns from three different product surfaces — web, iOS, and Android. Each surface team says fixing it is the other team's responsibility. How do you resolve this?

18.PyTorch is your training framework, and a model your team ships to production for content ranking on Facebook is performing well offline — AUC looks great — but online A/B metrics show a meaningful regression in meaningful engagement. What's your systematic approach to diagnosing the gap?

19.How would you design a system to detect and prevent coordinated inauthentic behavior on a social platform?

20.Meta's MySQL/MyRocks clusters are sharded by user_id, but a new privacy regulation requires you to implement a 'download your data' feature that must return every piece of content, activity, and metadata associated with a user within 30 days — across tables that are sharded differently or not by user_id at all. How do you architect this?

21.Facebook's comment ranking system has a silent regression — engagement metrics look fine, but qualitative research shows users feel the best comments are being buried. How do you diagnose this and what does fixing it look like end-to-end?

22.Meta is migrating a critical internal service from Hack to Rust for performance reasons. The service has 200+ callsites across 15 teams, no owner for many of the integrations, and needs to stay live during migration. How do you architect and lead this migration?

23.Design a content recommendation system for a social media platform with billions of posts.

24.System Design - delved a lot deeper

25.Tell me about a time when you had to balance short-term delivery pressure with long-term technical health.

26.Tell me about a time you identified that two or more teams were solving the same problem independently. How did you approach consolidating the work, and what resistance did you run into?

27.Describe a time you disagreed with a product direction that was already committed to — roadmap locked, stakeholders aligned, maybe even announced. What did you do?

28.You notice that a reliability metric your team owns — say, error rate or p99 latency — has been quietly degrading for months, but nobody escalated it because it never crossed the on-call threshold. How do you approach fixing not just the metric but the underlying process that let it drift?

29.Tell me about a time you led a significant technical decision that turned out to be wrong. Not a small bug — a real architectural or strategic call. How did you discover it was wrong, and what did you do about it?

30.You're told that WhatsApp's message open rate — the percentage of delivered messages that get opened within 5 minutes — dropped 8% last Tuesday and hasn't recovered. How do you figure out what's going on?

31.Instagram Reels watch-time is up 12% after a ranking model change, but shares and comments are flat and saves are down 7%. How do you interpret this, and would you ship the change?

32.Estimate how much server cost Meta incurs per day just from storing and serving Stories on Instagram — not video encoding, just storage and read serving. Walk me through your assumptions.

33.You discover that three separate Meta product teams — Marketplace, Groups, and Events — have each independently built their own location-based search infrastructure over the past two years. The systems are diverging, two of them are understaffed, and none of the teams want to give up ownership. How do you drive toward consolidation when you're a senior IC with no authority over any of the three teams?