Meta Backend Engineer Interview Questions

I would first ask how long the sessions need to live and whether they must survive server restarts. One option is to keep session state in a shared store such as a distributed cache or database. Another option is to use stateless tokens if the session data is small enough and does not need server-side invalidation every time. The reason I would compare both is that they solve different trade-offs between simplicity, revocation, and scalability. I would also mention token rotation, expiration, and what happens if the shared store becomes unavailable.

I would assume duplicates are possible and design the consumer to be idempotent. The consumer would store enough state to know whether a message has already been processed, such as a processed-message table or a deduplication key. The reason I choose this is that retries and re-delivery are normal in distributed messaging systems. I would also think about retention for deduplication state and how to handle partial work if a message fails midway. If the interviewer wants more depth, I would explain the difference between at-least-once and exactly-once processing.

I would add logs, metrics, and traces, but I would make sure they answer different questions. Metrics would show system health, logs would capture detailed events, and traces would show request flow across services. The reason I choose all three is that backend debugging usually needs both high-level trends and low-level context. I would also think about dashboards, alert thresholds, and keeping observability cost reasonable. If the interviewer asks, I would mention that bad observability often means missing correlation IDs or too much noisy logging.

I would route the notification through an internal event system first, then fan out to provider-specific delivery workers. That lets me handle email, push, and in-app notifications separately without coupling them tightly. The reason I choose this design is that each provider has different latency, retry, and failure behavior. I would also think about user preferences, deduplication, and dead-letter handling when a provider is unavailable. If the interviewer asks, I would mention batching and template versioning as well.

I would first find the bottleneck rather than assuming the answer is just more hardware. The main things I would inspect are the database, cache, queueing, connection limits, and whether the service can scale horizontally. The reason I start there is that traffic problems usually come from one constrained resource, not the whole application. I would also think about load shedding and graceful degradation so the system stays partially usable under stress. If the interviewer wants a concrete plan, I would say I would make the service stateless, protect the database, then add caching and queueing where needed.