System Design Case-Study Deep Dives¶
Thirty interview-grade system-design case studies, each broken into four files:
00-overview.md— problem, requirements, back-of-envelope scale math, API sketch, and the solutioning narrative with its defining tradeoffs.01-hld.md— architecture diagram, components, read/write data flow, storage choices, scaling, operational signals, and failure modes.02-lld.md— data models, component internals, core algorithms, a sequence diagram, and concurrency/edge cases.03-qa.md— interview Q&A (each with the common wrong answer to avoid), deeper follow-ups, and how the round is scored.
Every study threads one concrete numeric scenario through all four files. Authoring standard: _case_study_spec.md (excluded from the built site).
Core web-scale products¶
| # | Study | The scenario it turns on |
|---|---|---|
| 01 | URL Shortener | a campaign link at 30k redirects/sec |
| 02 | E-commerce Platform | a 10k-unit flash sale, 500k shoppers |
| 03 | a 200M-follower celebrity post | |
| 04 | Twitter/X Home Timeline | a tweet retweeted 1M times |
| 05 | WhatsApp / Messenger Chat | a message to an offline user, exactly-once |
| 06 | Video Streaming / OTT | 10M press play at 8pm |
| 07 | Live Streaming (Twitch) | 5M-viewer final + 50k-msg/s chat |
Marketplaces & transactions¶
| # | Study | The scenario it turns on |
|---|---|---|
| 08 | Payment System (Stripe) | a $50 charge retried 3× — charge once |
| 09 | Digital Wallet | two concurrent transfers, no overdraw |
| 10 | Ticketing System | 100k fans, 10k seats, no double-sell |
| 11 | Ride-hailing (Uber) | match a rider among 500 nearby drivers |
| 12 | Food Delivery | dinner rush, batch orders onto couriers |
| 13 | Hotel/Stay Booking | two guests, one room, overlapping dates |
Storage, search & collaboration¶
| # | Study | The scenario it turns on |
|---|---|---|
| 14 | File Sync (Dropbox) | a 1GB file edited on two devices |
| 15 | Collaborative Editor (Docs) | 50 people editing one paragraph |
| 16 | Search Autocomplete | 100k prefix queries/sec under 100ms |
| 17 | Web Crawler | crawl 10B pages, re-crawl news hourly |
| 18 | Proximity / Nearby Service | "restaurants within 1km" at 50k QPS |
Infra & platform building-blocks¶
| # | Study | The scenario it turns on |
|---|---|---|
| 19 | Rate Limiter | 100 req/s per key across 50 servers |
| 20 | Distributed Cache | a 100-node cluster loses a node |
| 21 | Message Queue (Kafka) | 1M events/s, ordered, day-behind catch-up |
| 22 | Notification System | 10M-subscriber breaking-news fanout |
| 23 | Monitoring / Metrics | 10M series, 1M points/sec ingest |
AI / ML systems¶
| # | Study | The scenario it turns on |
|---|---|---|
| 24 | Recommendation / Ranking | homepage recs for 200M users, <100ms |
| 25 | RAG Knowledge Assistant | Q&A over 10M docs with citations |
| 26 | Model-Serving / Inference | an LLM at 1k req/s under a p99 SLA |
| 27 | Vector Search Engine | 1B vectors, 10k QPS, 95% recall |
| 28 | Feature Store + Pipeline | no train/serve skew for a fraud model |
| 29 | Content Moderation / Fraud | 1M uploads/hour, ML + human review |
| 30 | LLM Agent Platform | a multi-step agent under a $1 budget |