Case studies with real architectures, plus the HLD and LLD concepts interviewers probe.
TinyURL/bit.ly — ID generation, redirects at scale, caching, and analytics.
Fan-out strategies, timeline caching, and the celebrity problem.
Geospatial indexing, real-time driver matching, and location update firehoses.
Video pipelines, adaptive bitrate streaming, and CDN-first architecture.
Message delivery guarantees, connection management, and end-to-end encryption plumbing.
Media-heavy feeds, blob storage with CDNs, and precomputed timelines.
Frontier queues, politeness, dedup at billions of pages, and the trap-filled real web.
Google-style search suggestions — tries with precomputed top-K, sub-50ms reads, and streaming updates.
Push, email, and SMS at scale — fan-out, rate limiting per user, dedup, and delivery guarantees.
File sync — chunking, content-addressed dedup, delta sync, and conflict resolution.
Seat booking under contention — reservations with TTL, virtual waiting rooms, and surviving the on-sale spike.
Real-time collaborative editing — OT vs CRDTs, presence, and why last-write-wins destroys documents.
How systems grow, why stateless services matter, and where the real bottlenecks hide.
L4 vs L7, balancing algorithms, health checks, and avoiding the LB as a single point of failure.
Cache-aside vs write-through, eviction, TTLs, and the three classic cache failure modes.
Partitioning strategies, shard keys, hot spots, and what you give up when you shard.
What CAP actually says, AP vs CP in practice, and the consistency spectrum beyond it.
Queues vs logs, delivery guarantees, idempotency, and when async is the right call.
Service boundaries, inter-service communication, and the costs nobody budgets for.
Edge caching, push vs pull, invalidation, and what belongs on a CDN.
Token bucket vs sliding window, distributed counters, and what to return when you throttle.
Why mod-N fails, the hash ring, virtual nodes, and where consistent hashing shows up.
Leader-follower, sync vs async, replication lag, and failover without losing writes.
Resource modeling, pagination, versioning, idempotency, and picking the right protocol.
Getting server events to clients — the four options, their costs, and scaling the connection layer.
Keeping multiple services consistent — 2PC, sagas, the outbox pattern, and idempotency as a way of life.
The three pillars, SLIs/SLOs and error budgets, and alerting that doesn't cry wolf.
The five principles with violations and fixes you can actually cite in an interview.
The dozen patterns that actually appear in LLD rounds, with when-to-use triggers.
A repeatable method for object-oriented design rounds — parking lot walked end to end.
The 20% of UML you need to whiteboard designs clearly — relationships, arrows, and multiplicity.
The canonical LLD warm-up — entities, enums vs hierarchies, spot allocation, and pricing strategies.
State machines, the SCAN scheduling insight, and dispatching across multiple cars.
Expense splitting — the balance-sheet model, split strategies, paise-rounding, and the debt-simplification graph problem.