M
Meta
Fully authored
2 rounds · 45 min each · either whiteboard
What they grade on
- Product-first framing — you connect the design to real user impact
- End-to-end coverage — from API to storage to failure modes in 45 minutes
- Explicit trade-offs — always compare at least two options
- Communication cadence — you drive the interview, not the interviewer
- Depth on 1–2 areas the interviewer probes
Deal breakers
- Silence when asked 'what would you do differently?'
- Refusing to make a decision
- No API before diving into architecture
- No consideration of privacy or abuse for a user-facing feature
Super helpful signals
- Naming a real Meta system or paper (TAO, F4, MyRocks, Scribe, Delos, Owl) as precedent
- A crisp abuse-prevention story for user-generated content
- Talking about impact on the news-feed ranking pipeline
Delivery advice
- Open with the product story: 'so a user opens Instagram, taps stories…' Ground everything in a user flow.
- Draw a small API (3–5 endpoints) before you draw architecture. The API is your contract with the interviewer.
- Every 10 minutes, pause and say 'here's what we've covered, here's what's next' — this earns communication points.
- When asked about scale, always give numbers before diagrams. 'Feed is read-heavy — 100:1 read:write, ~1M QPS peak' beats any picture.
- For any user-facing feature, sprinkle a privacy/abuse thought naturally.
Red flags to avoid
- Starting with microservices without justifying
- Silence on failure modes
- Vague answers to 'how does this scale?'
Top questions
Design News Feed ranking + delivery
Design real-time comments on a live video
Design Instagram Stories
Design a distributed key-value store
Design a metrics/monitoring service
Level expectations
L4 · E4
Correctness on one component; product intuition credit awarded
L5 · E5
End-to-end feature; explicit trade-offs
L6 · E6
Cross-team implications, capacity, ranking pipeline impact
L7 · E7
Precedent-setting, org design, platform proposals