Generated by Codex with gpt-5

Selected problem: Search Autocomplete

Scope: Design a low-latency autocomplete service that returns top suggestions for a typed prefix, learns from query activity over time, and handles hot prefixes, moderation, and multilingual growth without rebuilding everything on every keystroke.

Problem framing

This is the classic “design Google search suggestions” or “typeahead” interview problem. Alex Xu frames it around a brutally simple user-facing requirement: every keystroke can trigger a request, so suggestion reads must be extremely fast. Grokking’s interview style also applies cleanly here: clarify prefix-only versus infix matching, top-K size, language scope, freshness needs, and latency targets before arguing about data structures. The interview answer usually starts with a trie or prefix tree, but DDIA adds the more durable framing: autocomplete is a derived read model built from query logs, aggregation, filtering, and ranking pipelines, so the real design question is how to keep that read model fresh, cheap, and rebuildable.

Generated by Codex with gpt-5

Selected problem: Web Crawler

Scope: Design a large-scale, HTML-first web crawler for search indexing that discovers new URLs, respects host politeness and robots rules, deduplicates content, and keeps important pages reasonably fresh.

Also see https://wiki.derricklin.net/software-development/System%20Design%20Interview/#web-crawler

Problem framing

This is the classic “design Googlebot” interview question. Grokking emphasizes the minimum crawler loop: start from seed URLs, fetch pages, extract links, dedupe, and repeat. Alex Xu frames the real interview difficulty more accurately: the hard parts are URL frontier design, politeness, prioritization, freshness, robustness, and avoiding bad content. DDIA supplies the missing systems lens: frontier queues, dedupe tables, document storage, and downstream indexing are separate dataflow stages, so the design should favor partitioned state, replayable logs, and idempotent consumers rather than pretending the entire pipeline is one giant transaction.

Generated by Codex with GPT-5

Quick facts

• Difficulty: HARD
• Problem: Maximal Rectangle
• Main tags: Array, Dynamic Programming, Stack, Matrix, Monotonic Stack

What the problem is really asking

The matrix contains only "0" and "1", and the goal is to find the area of the largest axis-aligned rectangle made entirely of "1" cells.

Generated by Codex with GPT-5.4

Quick facts

• Difficulty: MEDIUM
• Problem: Rotate Array
• Main tags: Array, Math, Two Pointers

What the problem is really asking

The problem gives an array nums and an integer k. The task is to rotate the array to the right by k positions, modifying the original list in place.

Generated by Codex with GPT-5

Quick facts

• Difficulty: MEDIUM
• Problem: Course Schedule
• Main tags: Depth-First Search, Breadth-First Search, Graph Theory, Topological Sort

What the problem is really asking

There are numCourses courses labeled from 0 to numCourses - 1.

Each prerequisite pair [a, b] means course b must be finished before course a.

Generated by Codex with GPT-5

Quick facts

• Difficulty: MEDIUM
• Problem: Subsets
• Main tags: Array, Backtracking, Bit Manipulation

What the problem is really asking

Given an array nums of distinct integers, return every possible subset of those numbers.

A subset can contain none of the numbers, some of the numbers, or all of the numbers. The empty subset [] is valid, and the full array is valid too. Because the input numbers are distinct, there is no duplicate-handling problem in the base version.

Generated by Codex with gpt-5

Selected problem: API Rate Limiter

Scope: Design a distributed server-side rate limiter that enforces per-user, per-IP, per-tenant, and per-route API quotas with very low decision latency across many stateless services.

Problem framing

This is the classic server-side API rate limiter from Grokking and Alex Xu: every request should get a cheap admission decision before it reaches expensive application logic.

Generated by Codex with gpt-5

Selected problem: Distributed Cache

Scope: Design a distributed in-memory cache service that provides very low-latency GET/SET/DELETE/INCR operations with TTLs, eviction, sharding, and failover for many stateless application services.

Problem framing

This is the classic “design Memcache / Redis-class cache” interview problem: build a shared cache tier that is much faster than the system of record, scales horizontally, and fails gracefully instead of turning into a new bottleneck.

Generated by Codex with gpt-5

Selected problem: Notification System

Scope: Design a soft real-time notification platform that accepts events from internal services, renders user-facing messages, respects user preferences, and dispatches push, email, SMS, and in-app notifications reliably at large scale.

Also see https://wiki.derricklin.net/software-development/System%20Design%20Interview/#notification-system

Problem framing

This is the classic “design a notification system” interview problem: build the shared infrastructure that product teams use whenever they need to notify users about security alerts, messages, payments, delivery updates, marketing campaigns, or feed changes. The core challenge is not just calling APNs, FCM, an SMS gateway, or an email provider. The harder parts are durable intake, preference filtering, fanout, queue isolation, retries, deduplication, rate limiting, observability, and honest delivery semantics.

Generated by Codex with GPT-5

Quick facts

• Difficulty: HARD
• Problem: Largest Rectangle in Histogram
• Main tags: Array, Stack, Monotonic Stack

What the problem is really asking

Each bar in the histogram has width 1, and the goal is to find the largest possible rectangle that can be formed by taking one or more adjacent bars.