Backend Engineer Interview Questions
Practice API, database, distributed-system, reliability, and execution questions for backend engineering roles.
Best for: Backend engineers from entry through senior level.
Questions and what strong answers sound like
Answer each one aloud first, then compare.
-
How would you make a payment-creation endpoint idempotent?
Strong-answer signals: Take an idempotency key from the client, store it with the request fingerprint and the response, and return the stored response on a retry. Say how long you keep keys, and what happens when two identical requests arrive at the same moment (a unique constraint or a lock on the key).
-
When would you denormalize a database model?
Strong-answer signals: Start from the slow queries you measured, not from a rule. Say what you gain on reads, what it costs on writes, how you would repair drift, and which table stays the source of truth.
-
A queue consumer falls hours behind. How do you diagnose and recover?
Strong-answer signals: Compare the arrival rate with what the consumers actually process, then find what is slow: one poison message, a slow database call, or too few partitions. Scale only after you know the dependency can take it, and replay the backlog at a rate that does not knock it over.
-
Your API has a 200ms p50 but a 4s p99. Where do you look?
Strong-answer signals: Averages hide this, so split p99 by endpoint and by downstream call. Typical culprits are lock waits, connection pool exhaustion, GC pauses and retries stacking up. Fix the one you measured before adding timeouts or shedding load.
-
How would you design rate limiting for a public API with paid tiers?
Strong-answer signals: Compare token bucket and sliding window for burst behavior, then decide where the counters live. Redis gives shared limits but is a dependency, so say whether you fail open or closed when it is down. Return 429 with Retry-After and limit per API key.
-
Two services must update their own databases as one business operation. What do you do?
Strong-answer signals: Skip two-phase commit unless you must. Write the change and an outbox row in one local transaction, publish from the outbox, and use a saga with compensating steps for failures. Consumers have to be idempotent, and users may see intermediate states.
-
You need to add a NOT NULL column to a table with 500 million rows. How do you ship it?
Strong-answer signals: Add the column as nullable, backfill in small batches while watching locks and replication lag, then add the constraint. Deploy code that handles both states first, so every step can be rolled back.
-
How do you choose between a cache, a read replica and a better index?
Strong-answer signals: Read the query plan first: a missing index is the cheapest fix. A cache helps hot, repeated reads but you pay in invalidation and staleness. A replica adds capacity, with lag that breaks read-your-writes.
-
A downstream service starts failing intermittently. How does your service behave?
Strong-answer signals: Set a timeout shorter than the caller's, retry a bounded number of times with jittered backoff, and stop retrying when the dependency is clearly down (a circuit breaker). Decide what the user sees instead, and alert on the dependency's error rate.
-
Tell me about a backend decision you reversed.
Strong-answer signals: Pick a real decision. Explain what you knew when you made it, what evidence changed your mind, how you migrated, and how you told the people affected.
Interviewers ask about what is on your resume.
Upload yours and get feedback on how the projects read.
Frequently asked questions
What should I study for a Backend Engineer interview?
Prioritize API contracts, Data modeling, Consistency, Reliability, Delivery judgment. Connect each concept to production decisions, failure modes, and work you can explain from direct experience.
How should I use the strong-answer signals?
Use them as a self-review checklist after answering aloud. They are not a script and strong solutions can take different approaches when their tradeoffs are explicit.
Should I memorize these answers?
No. Practice a repeatable reasoning process: clarify requirements, state assumptions, compare options, choose deliberately, and explain how you would verify the result.