Interview questions with answer breakdowns — Page 48 of 52
Browse page 48 of Offer.cc interview question and answer breakdowns with reasoning, implementation details, follow-ups, and public sources.
Backend Interview: How Should Timeouts, Retries, and Circuit Breakers Work Together?
Design synchronous dependency calls with an end-to-end deadline, a failure-semantics matrix, and a retry budget, then contain cascading failures with circuit breakers, bounded probes, and fault injection.
Data Engineering Interview: How Do You Safely Backfill Large-Scale Historical Data?
Split a 180 TB recomputation into recoverable partitions, then protect the live pipeline with deterministic inputs, isolated resources, delta catch-up, validation gates, and a versioned cutover.
Coding interview: Implement a work-stealing scheduler
Build per-worker deques with local LIFO and remote FIFO stealing, then reason about shutdown, idleness, fairness, and duplicate execution.
System Design Interview: How Would You Design a Personalized News Feed?
Design a personalized news feed around candidate generation, hybrid fan-out, staged ranking, stable pagination, authorization filtering, and recovery from celebrity hotspots and eventual consistency.
Backend Interview: How Do You Diagnose and Fix the N+1 Query Problem?
Detect query counts that grow linearly with page size, compare JOIN and select-in batch loading, and verify the fix without hiding overfetching or consistency changes.
Data Engineering Interview: How Do You Diagnose and Fix Spark Data Skew?
Prove Spark data skew with the UI, execution plan, and key distribution, then choose AQE, broadcast, targeted join salting, or two-stage aggregation and verify both correctness and performance.
Coding interview: Implement per-key request coalescing (singleflight)
Implement in-flight deduplication by key with shared results, retry after failure, race-safe cleanup, and explicit cancellation semantics.
System Design Interview: Design Consistent Hashing with Virtual Nodes
Derive a hash ring from modulo remapping, then design consistent hashing with virtual nodes, weights, replicas, and versioned migration while identifying the hot-key and failure problems it cannot solve.
Backend Interview: Explain OAuth Authorization Code Flow with PKCE
Derive the OAuth authorization code flow across authorization, callback, and token exchange; separate state, PKCE, client authentication, and OIDC; then verify the design with attacks and failure tests.
SQL Interview: Calculate Exact Day-7 Cohort Retention
Build an exact Day-7 cohort-retention query in PostgreSQL, with calendar-day time zones, mature-cohort filtering, user-level deduplication, zero-return preservation, and production checks.
Coding interview: How would you design an error-recovering expression parser?
Build lexing, parsing, synchronization, diagnostics, and partial ASTs so one syntax error does not invalidate an entire document.
System Design Interview: Design an E-commerce Payment Processing System
Derive a recoverable e-commerce payment system from payment intents, authorization and capture, idempotent retries, asynchronous webhooks, an immutable money ledger, refunds, and reconciliation.
Backend Interview: How Do You Prevent SQL Injection?
Separate SQL code from data with server-side parameterization, map dynamic identifiers through fixed allowlists, preserve tenant authorization, and verify raw-query paths, stored procedures, and second-order execution.
SQL Interview: Find Each User's Longest Consecutive Login Streak
Solve a PostgreSQL gaps-and-islands interview problem by fixing the business-day grain, deriving a stable group key with ROW_NUMBER, preserving tied longest streaks, and testing time-zone and duplicate-event edge cases.
Coding Interview: How do you implement a cancellable priority task scheduler?
Implement a single-node scheduler with deadlines, priorities, duplicate-task replacement, and cancellation while handling stale heap entries, worker limits, and shutdown races.
System Design Interview: Design a URL Shortener
Design a durable, low-latency URL shortener by deriving code-space, storage, cache, redirect, idempotency, expiration, and abuse controls from explicit traffic assumptions.
Backend Interview: How Do You Size a Database Connection Pool and Diagnose Exhaustion?
Derive connection-pool limits from a database-wide budget, request concurrency, and replica count, then distinguish an undersized pool, a leak, and database overload with application queues and PostgreSQL session evidence.
Data Engineering Interview: How Do You Debug Wrong Dashboard Numbers?
Start with metric semantics and a trustworthy comparator, trace lineage to the first bad boundary, then restore trust with containment, idempotent backfills, reconciliation, and data contracts.
Coding interview: Implement a concurrency-limited async map with cancellation
Use fixed workers, indexed results, explicit error modes, and AbortSignal to solve an async scheduler problem.
System Design Interview: Design a Distributed Message Queue
Derive a replayable, scalable, multi-tenant message queue from delivery contracts, partitioned logs, replica acknowledgments, consumer offsets, and an object-storage path for large payloads.
How Do You Perform a Zero-Downtime PostgreSQL Schema Migration?
Start with rolling-deployment compatibility, then design an abortable expand-backfill-cutover-contract workflow that handles PostgreSQL DDL locks, dual writes, constraint validation, indexes, and phase-specific rollback.
Data Engineering Interview: How Do You Design an ML Feature Pipeline Without Training-Serving Skew?
Using real-time fraud scoring, design a verifiable ML feature pipeline with a semantic contract, two clocks, offline and online stores, versioned rollout, and feature-vector replay.
Coding Interview: Implement a Time-Versioned Key-Value Store
Implement set/get with per-key version histories and upper-bound binary search, covering out-of-order timestamps, ties, boundaries, and memory trade-offs.
System Design Interview: How Would You Design a Metrics Monitoring and Alerting System?
Start with time-series and cardinality budgets, then design a metrics platform with high-throughput collection, bounded queries, reliable alerts, tenant isolation, and testable recovery.