67. Interview Prompt Bank: Real-World Questions, Mapped

These are system design prompts commonly reported in senior loops (Senior, Staff, Principal, and engineering managers) at tier-1 companies, grouped by category. For each one: what it really tests, the 2–3 decisions that make or break the answer, and where this guide covers it. Practice at least two from each category.

67.1 How to approach a prompt you've never seen (5 moves)

 1. Name the business & the user  → "Who pays, what outcome, what does failure cost?"
 2. Find the dominant force       → read-heavy? write-heavy? latency? correctness? cost? safety?
 3. Find the hard key             → what is the partition key / unit of consistency / unit of isolation?
 4. Find the one-way doors        → data model, consistency promise, public API, trust model
 5. Map to known patterns          → use §68.1 (problem → pattern map) and the casebook below

67.2 Consumer products

Prompt	What it really tests	Key decisions	Covered in
Design Instagram / photo sharing	Breadth, scoping, media cost, feed	Hybrid fan-out; media tiers + CDN; counters	§48, §33
Design Twitter/X timeline	Fan-out trade-offs, celebrity problem	Push vs pull hybrid; ranking budget	§33
Design TikTok / Reels "For You" feed	Recommendation funnel, video delivery	Candidate generation → ranking; prefetching; cold start	§25.1, §37
Design YouTube / Netflix	Upload pipeline, ABR, CDN economics	Chunked parallel transcode; per-title ladders; ISP caches	§37, §49
Design WhatsApp / Messenger	Ordering, delivery guarantees, connections	Per-conversation sequence; durable before ack; gateway tier	§34
Design Uber / Lyft	Geo index, matching, real-time state	H3 in memory; atomic driver offer; batch matching	§36
Design Yelp / "nearby places"	Read-heavy geo search	Geohash/quadtree; cache by cell; ranking	§15.2
Design Google Maps / navigation	Graph routing at scale, live traffic	CRP/contraction hierarchies; traffic as a metric update; vector tiles	§51, §15.2, §36
Design Airbnb / hotel booking	Search + inventory correctness	Availability calendar consistency; holds; search index freshness	§40, §15
Design Ticketmaster	Contention + spikes + fairness	Waiting room; conditional holds; bot defense	§40
Design Spotify	Catalog, streaming, offline sync, recommendations	CDN for audio; license/region rules; offline DRM	§37, §25
Design Zoom / Teams meetings	Real-time media (SFU vs MCU), signaling	SFU fan-out; regional media servers; jitter buffers	§5.1, §20
Design Google Docs / Office co-authoring	Concurrency control	OT vs CRDT; single owner per doc; op log + snapshots	§41
Design Dropbox / OneDrive / Drive	Sync correctness, dedupe, durability	Content-defined chunking; journal cursors; erasure coding	§45
Design Gmail / Outlook	Huge per-user storage, search, spam	Per-user sharding; per-mailbox index; anti-abuse pipeline	§15, §17
Design a news aggregator (Google News)	Crawl + cluster + rank + freshness	Dedupe/clustering; freshness scoring	§43, §25
Design a leaderboard (gaming)	Real-time ranking at scale	Sorted sets sharded by score range; approximate ranks for the tail	§7, §9

67.3 Infrastructure & platform

Prompt	What it really tests	Key decisions	Covered in
Design a service mesh	Control/data plane, config fan-out, identity	Scoped delta xDS; L4 node proxy vs sidecar; static stability	§57
Design a CDN	Routing, caching hierarchy, global purge	Anycast; PoP-internal hashing; generation-based purge	§49
Design a certificate authority / PKI	Trust bootstrapping, expiry, rotation	Attestation; offline root; 24 h certs + renew at 50%	§56
Design a secrets manager / KMS	Envelope encryption, access policy, audit	Key hierarchy (KEK/DEK); HSM roots; lease-based secrets	§55, §17, §56
Design a distributed rate limiter	Accuracy vs latency, multi-region	Local + global hybrid; fail-open policy	§32
Design a distributed KV store	Partitioning, replication, consistency	Consistent hashing + quorums, or Raft per range	§38
Design a distributed cache (Memcached/Redis as a service)	Hot keys, invalidation, failure behavior	Leases; replication of hot keys; gutter pool	§7
Design a message queue / Kafka	Log storage, replication, consumer groups	ISR + acks; partitioning; retention	§11
Design object storage (S3)	Durability math, metadata scale	Erasure coding; metadata partitioning; consistency of listings	§52, §45, §6
Design a load balancer	L4 vs L7, consistent hashing, health	Maglev-style hashing; connection draining	§4.5
Design DNS / service discovery	Caching, TTLs, propagation	Anycast resolvers; TTL vs failover speed	§4, §19
Design a job scheduler / cron	Exactly-once launch, leases	Unique execution records; fencing	§46
Design a metrics/monitoring system	Cardinality, compression, alert isolation	Head + blocks; isolated alert path	§53
Design a logging pipeline	Volume, sampling, cost	Agent buffering; tiered storage; index-light designs	§16, §21
Design distributed tracing	Sampling, context propagation	Tail-based sampling; trace ID everywhere	§16
Design a feature-flag / config system	Safe change at scale	Local evaluation; staged config; static stability	§54
Design a CI/CD system	Scheduling, caching, safe rollout	Hermetic builds + remote cache; progressive delivery	§29.4
Design a container orchestrator (Kubernetes/Borg)	Scheduling, reconciliation loops	Desired-state controllers; etcd; bin packing	§58, §19, K8s guide
Design an API gateway	Routing, auth, limits, multi-tenancy	Thin gateway; per-tenant limits; config distribution	§13.3, §32
Design an identity / SSO service	Token issuance, sessions, federation	OIDC; short tokens + refresh; revocation	§17
Design an authorization service (Zanzibar)	Relationship ACLs, consistency	Tuples; zookies; caching + hedging	§17.2

67.4 Data, ML & search

Prompt	What it really tests	Key decisions	Covered in
Design search autocomplete	Latency, precomputation, policy	Trie with top-K per node; trending overlay	§42
Design a web crawler	Politeness, dedupe, prioritization	Two-level frontier; SimHash; recrawl scheduling	§43
Design web search (Google)	Indexing, sharding, ranking	Doc-partitioned index; tiered serving; multi-stage ranking	§50, §15, §43
Design ad click aggregation	Exactly-once counting, late data	Dedupe + watermarks; batch reconciliation	§44
Design top-K / trending	Streaming aggregation, approximation	Windows + heap; Count-Min Sketch	§44
Design a recommendation system	Funnel, features, feedback loops	Embeddings retrieval; feature store; online metrics	§25.1
Design a fraud detection system	Real-time scoring with fallbacks	Feature freshness; fail-open vs fail-closed by risk	§39, §25
Design a RAG assistant / LLM app	Retrieval with ACLs, evals, cost	Hybrid search; ACL filtering in index; eval CI	§25.2
Design an AI agent platform / coding agent	Autonomy vs control, tool security, durability	Workflow vs agent; sandboxed tools + approvals; durable runtime + evals	§26
Design an LLM serving platform	GPU efficiency, latency SLOs	Continuous batching; KV cache; routing by SLO tier	§25.3
Design a data warehouse / lakehouse	Batch + streaming, governance	CDC ingestion; table formats; data contracts	§21
Design a feature store	Online/offline consistency	Shared definitions; point-in-time correctness	§25.1

67.5 Fintech & commerce

Prompt	What it really tests	Key decisions	Covered in
Design a payment system (Stripe)	Idempotency, state machines, reconciliation	Unknown-outcome handling; double-entry ledger	§39
Design a digital wallet / ledger	Correctness under concurrency	Immutable entries; hot-account splitting	§39
Design a stock exchange	Determinism, latency, fairness	Sequencer; single-threaded shards; replay	§59
Design an e-commerce checkout	Inventory + payment saga	Reservations with TTL; saga compensation	§12, §40
Design a flash-sale system	Contention, admission control	Waiting room; sharded inventory counters	§40
Design a currency exchange / FX service	Rate freshness vs consistency	Rate snapshots with validity; quote locking	§39
Design a billing / subscription system	Proration, retries, dunning	Event-sourced invoices; idempotent charges	§39, §46

67.6 Enterprise & multi-tenant SaaS (frequent at Microsoft, Google Cloud, AWS)

Prompt	What it really tests	Key decisions	Covered in
Design a multi-tenant SaaS platform	Isolation vs cost, noisy neighbors	Tiered tenancy; cells; tenant directory	§17, §20.4
Design an audit logging service	Immutability, query, retention	Append-only + Merkle proofs; tiered retention	§17, §56
Design Slack / Teams	Channels, presence, enterprise compliance	Channel servers; retention/legal hold; tenant cells	§34
Design a calendar service	Recurrence, time zones, free/busy	RRULE expansion; time-zone correctness; sharing ACLs	§46 corner cases
Design a notification platform	Priorities, preferences, providers	Transactional vs marketing isolation; budgets	§35
Design an email delivery service (SES/SendGrid)	Reputation, throughput, bounces	IP warm-up/pools; per-tenant reputation; suppression lists	§35
Design a document/e-signature workflow (DocuSign)	Workflow durability, legal audit	Orchestrated workflow; tamper-evident audit	§12, §46

67.7 Leadership "meta" prompts for Staff+, Principal, and EMs (no single right architecture)

Prompt	What they want to hear	Covered in
"Our monolith can't keep up. Plan the next 2 years."	Diagnose first; modular monolith vs extraction; strangler; staffing; milestones	§13, §29, §30
"Migrate our database to X with zero downtime."	Expand/migrate/contract; dual-write vs CDC; verification; rollback	§29
"Cut infrastructure cost by 30% without hurting SLOs."	Unit economics; top cost lines; levers ranked by effort/impact	§28
"We have frequent outages. What do you do in your first 90 days?"	Incident data → top causes (usually change management); SLOs; error budgets; staged rollout	§14, §16, Appendix A
"Make our product available in the EU with data residency."	Region pinning; tenant routing; DR within the boundary; deletion workflows	§17, §20
"Design the platform 300 engineers will build on."	Paved road; self-serve; ownership; adoption metrics	§13, §30, §57, §54