Updated by Meta candidates
Senior Software Engineer, Infrastructure Interview Experience
There’s a new AI coding section I haven’t seen before. They still want to test my coding skill, so I first implement by myself, then use AI as a code reviewer for readability and missing test cases.
Interview process
The process was first a phone screen followed by a virtual loop with four rounds. The phone screen was a single tree traversal coding problem, and the onsite had two coding rounds, one system design round, and one behavioral round. The most interesting part was a new AI coding section that I hadn't seen before. It sounded new enough that even the format felt a little flexible, but the expectation was still that I do the actual coding myself and only use AI in a limited way. Overall it was pretty challenging, and if I did it again I would spend more time on behavioral prep because that was my weakest part.
- Phone interview
- Final round
Interview tips
I would focus hard on coding speed and correctness because the coding questions felt very practiceable. If you do enough LeetCode-style prep, you should be able to find very similar problems, so make sure you can implement them fast and clean. I would also prepare behavioral stories much more than I did, especially stories you can map to Meta's values. And if you're not sure whether the role is more product or infra, ask the recruiter, because the system design emphasis is different.
Company culture
I got the sense they're experimenting a bit with the process right now because the AI coding round felt new and didn't have a super concrete format yet. The interviewer was still clear that AI is not a shortcut and they want to see my own coding skill first.
Questions asked
Overview
My virtual loop had four rounds: two coding rounds, one system design, and one behavioral. The standout was a new AI coding round I hadn't seen before, and the interviewer said they still wanted to test my coding skill, so I couldn't just let AI do everything. The infra system design was also noticeably more database and scalability focused than a product-style design interview.
Question types asked
Specific questions asked
They gave me a preorder traversal and asked me to reconstruct the BST. I solved it recursively and focused on building the tree back correctly from the preorder input.
The second question in that coding round was to compute x^n without using any built-in library. I just had to implement it directly and get it correct.
How are you supposed to use the AI tool during this round?
What did they say about the format and what they were evaluating?
This round felt more like object-oriented design. I implemented the classes and the required functionality like shuffling cards and assigning cards to players. The interviewer said they still wanted to test my coding skill, so I wasn't supposed to rely on AI for the whole implementation. My strategy was to code it myself first, then use AI more like a code reviewer for readability improvements and to catch missing test cases. The format felt pretty new and flexible.
What transport would you use for real-time updates: WebSocket, server-sent events, or long polling?
How would you support fan-out?
How would you think about SQL vs NoSQL for persistence?
I designed the whole system including APIs, database, and the overall architecture for millions or billions of users. A lot of the discussion was around real-time delivery and which technology to use, like WebSocket vs server-sent events vs long polling. I compared the pros and cons and ended up choosing WebSocket. We also talked about using a pub-sub system for fan-out and I compared SQL and NoSQL for persistence. I felt like this round really emphasized database and infrastructure tradeoffs.
Why was it challenging?
What was the scale?
Did you work with other teams?
This one was more freestyle. I spent a couple minutes talking through a project I was proud of, then answered follow-ups on why it was challenging, what kind of scale it had, and how I worked with other teams. In hindsight, I don't think I had a really strong story prepared here.
I answered it, but honestly this was another place where I felt underprepared. Afterward I realized I should have prepared better behavioral stories and tied them more clearly to Meta's values.
Get full access with a membership, or share your experience to try it free.
