System designLast reported Aug 2026
Design a Gang-Scheduled GPU Inference Pool
Design an inference request path whose cache misses become one-GPU or all-GPU batches competing for a fixed accelerator pool.
Open practiceAnthropic interview practice
Anthropic interview questions and coding practice. Start with verified evidence.
Practice from 21 Anthropic-tagged public coding problems, organized only by the stage, topic, difficulty, and recency metadata FastPrep can verify. The broader public catalog also includes 7 system-design exercises, 3 project-coding exercises.
- Catalog
- 31 public practice assets (21 coding)
- Start here
- Use the stage named in your invitation
- Evidence
- OA · Phone · Onsite metadata, shown in context
OA practice path
Online Assessment
Build a timed implementation rhythm from prompt to edge cases. Problems are ranked by repeated public catalog sightings.
17 Anthropic-tagged Online Assessment problems available.
| Company | Problem | Difficulty | Public evidence | Action |
|---|---|---|---|---|
ANAnthropic | In-Memory Database with Backup and RestoreDesignHash Table | Hard | 13 public reportsLast reported May 2026 | Practice |
ANAnthropic | Banking System, Part 1: Accounts and TransfersDesignHash Table | Easy | 10 public reportsLast reported Jul 2026 | Practice |
ANAnthropic | Banking System, Part 2: Top SpendersDesignHash Table | Medium | 10 public reportsLast reported Jul 2026 | Practice |
ANAnthropic | Banking System, Part 3: Scheduled PaymentsDesignHash Table | Hard | 10 public reportsLast reported Jul 2026 | Practice |
ANAnthropic | Banking System, Part 4: Merging and Balance HistoryDesignHash Table | Hard | 10 public reportsLast reported Jul 2026 | Practice |
ANAnthropic | Cloud Storage SystemDesignHash Table | Medium | 9 public reportsLast reported Aug 2026 | Practice |
ANAnthropic | In-Memory Database Historical LookupHash TableDatabase | Hard | 6 public reportsLast reported Nov 2025 | Practice |
ANAnthropic | Worker Management, Part 1: Office RegistrationHash TableDesign | Easy | 4 public reportsLast reported May 2026 | Practice |
ANAnthropic | Worker Management, Part 2: Top WorkersHash TableSorting | Medium | 4 public reportsLast reported May 2026 | Practice |
ANAnthropic | Worker Management, Part 3: Promotions and SalaryHash TableIntervals | Hard | 4 public reportsLast reported May 2026 | Practice |
ANAnthropic | Worker Management, Part 4: Double-Paid IntervalsHash TableIntervals | Hard | 4 public reportsLast reported May 2026 | Practice |
ANAnthropic | Cloud Storage Compression and DecompressionHash TableSorting | Hard | 3 public reportsLast reported Feb 2026 | Practice |
ANAnthropic | Banking Pending Transfer AcceptanceHash TableDesign | Medium | 2 public reportsLast reported Jul 2024 | Practice |
ANAnthropic | Task Management SystemDesignHash Table | Hard | 1 public reportLast reported Aug 2026 | Practice |
ANAnthropic | Normalize a DNS Domain NameString | Easy | 1 public reportLast reported May 2026 | Practice |
ANAnthropic | Cloud Storage File VersioningHash TableSorting | Medium | 1 public reportLast reported Apr 2024 | Practice |
ANAnthropic | Cloud Storage Prefix Trash and RestoreHash TableString | Hard | 1 public reportLast reported Apr 2024 | Practice |
Phone practice path
Phone Screen
Practice solving while explaining assumptions, tradeoffs, and complexity. Problems are ranked by repeated public catalog sightings.
| Company | Problem | Difficulty | Public evidence | Action |
|---|---|---|---|---|
ANAnthropic | Convert Stack Samples to Trace EventsStackSimulation | Medium | 11 public reportsLast reported Jul 2026 | Practice |
ANAnthropic | Profiling Events from Consecutive Stack SamplesArrayStack | Hard | 2 public reportsLast reported Aug 2026 | Practice |
Onsite practice path
Onsite
Practice complete solutions and defend the choices behind them. Problems are ranked by repeated public catalog sightings.
| Company | Problem | Difficulty | Public evidence | Action |
|---|---|---|---|---|
ANAnthropic | Repair the Bootloader ProgramArraySimulation | Medium | 2 public reportsLast reported Jun 2026 | Practice |
ANAnthropic | Cache Shared Image Pipeline PrefixesArrayMatrix | Hard | 1 public reportLast reported Aug 2026 | Practice |
01 · Stage workflow
Practice the stage you actually have next.
Combine algorithm work with public system-design and project-coding practice so preparation covers both narrow implementation and larger engineering decisions. Your invitation and recruiter guidance remain the source of truth.
- 01
Online Assessment
Use the timing and format in your invitation as the source of truth. In practice, work without interruptions, budget time to read constraints, and reserve a final review window for boundary cases.
- State the input, output, and complexity target before coding.
- Implement one complete solution inside a fixed timebox.
- Run small, duplicate, empty, and boundary cases before finishing.
- 02
Phone Screen
Treat the problem as a conversation. Clarify the contract, describe the approach before implementation, and keep narrating meaningful decisions while you code and test.
- Ask clarifying questions instead of silently filling gaps.
- Explain the data structure and expected complexity.
- Walk through at least one example and one edge case aloud.
- 03
Onsite
Use the catalog's onsite labels as practice context, not as a promise about a specific loop. Focus on collaborative problem solving: compare approaches, implement cleanly, and explain how the solution behaves at scale.
- Compare a straightforward approach with the optimized one.
- Keep the implementation readable while discussing tradeoffs.
- Test the result, then name follow-up improvements explicitly.
02 · Broader technical practice
Practice beyond standalone coding questions.
These public Anthropic-tagged exercises cover additional technical formats. They are included only when a verified catalog record and a crawlable practice page both exist.
System designLast reported Aug 2026
Design Opaque-GPU Inference Capacity
Size and route online and offline inference across opaque GPU workers using measured capacity, batching, and explicit overload policy.
Open practiceSystem designLast reported Aug 2026
Design a Model-Weight Deployment Platform
Design a control plane that safely distributes large immutable model versions to serving fleets through staged rollout, verification, activation, and rollback.
Open practiceSystem designLast reported Aug 2026
One-to-One Chat Service
Design a durable real-time messaging service for private conversations between two users, including ordering, offline recovery, and failure handling.
Open practiceSystem designLast reported Aug 2026
Design Peer-to-Peer Model Distribution
Design a topology-independent peer-to-peer system that disseminates one large immutable model while respecting a shared per-node bandwidth limit.
Open practiceProject codingLast reported Aug 2026
Debug Extremely Randomized Trees (Fellow Program :)
Debug the provided Extremely Randomized Trees implementation so it is correct and no longer suffers bugs that cause crashes or poor accuracy.
Open practiceProject codingLast reported Aug 2026
Inference Engine (Fellow Program :D
Implement the request scheduler of an LLM inference engine, assigning prefill and decode work across progressively harder levels.
Open practiceSystem designLast reported Aug 2026
Design Production ML Serving Observability
Scale and optimize an existing online ML service using end-to-end metrics, traces, logs, controlled releases, and explicit overload behavior.
Open practiceSystem designLast reported Aug 2026
Design a Prompt Playground
Design a collaborative prompt experimentation product with immutable versions, reproducible runs, searchable results, and safe model-provider execution.
Open practiceProject coding
Repair an Agent and Reduce Its Turn Count
Repair a supplied agent, make it answer every question within two turns, then reduce the solution to one turn.
Open practice03 · Evidence boundary
What “Anthropic interview questions” means here.
It means practicing transferable implementation, testing, and technical reasoning with public catalog assets FastPrep tags to Anthropic. It does not mean FastPrep has access to the company's assessments or any private interview bank.
Repeated public sightings and last-reported dates can help you prioritize practice, but they cannot predict the questions, format, or platform in a specific interview.
- Only Anthropic-tagged FastPrep catalog problems appear above.
- Stage labels and report dates remain visible as context.
- No official, private, leaked, or proprietary questions are claimed.
04 · Plain answers
Anthropic interview practice questions, answered plainly.
This page owns technical-practice intent for Anthropic. Hiring activity, timelines, and market signals remain on the separate hiring-insights page.
01Does this page include Anthropic design or project practice?
Yes, when the public catalog supports it. This page currently links 7 system-design exercises, 0 low-level-design exercises, and 3 project-coding exercises tagged to Anthropic. Empty formats are omitted rather than padded with generic recommendations.
02Is FastPrep affiliated with Anthropic?
No. FastPrep is an independent interview-preparation product and is not affiliated with Anthropic. The page uses FastPrep's public practice catalog and does not claim official, private, leaked, or proprietary employer questions.
03How are Anthropic problems assigned to interview stages?
Stage labels come from FastPrep's public problem metadata. A problem can carry more than one reported stage, and hiring processes can change by role, level, location, and date. Treat the labels as preparation context, not a guarantee.
04Which Anthropic interview stage should I practice first?
Start with the stage named in your invitation or recruiter message. If no stage is known, use the largest available set to build general problem-solving fluency, then rehearse explanation and testing separately.
05Does this page predict the Anthropic interview process?
No. Public catalog counts, stage tags, and last-reported dates can help prioritize practice, but they cannot predict a specific interview's questions, sequence, timing, or platform.
Your next stage, made concrete
Pick a problem. Solve it end to end.
Start with the stage named in your invitation, then use public evidence as context—not as a promise of what you will be asked.
Open the practice set