Table of contents

• Where do I apply for an Anthropic interview?
• What topics should I study for?
• What are some example Anthropic technical questions?
• What are some example Anthropic behavioral questions?
• How long is the entire Anthropic interview process?
• What is the Anthropic interview like for a full time software engineering role?
• What are the levels and salary for a software engineer at Anthropic?
• How should I prepare?

Where do I apply for an Anthropic interview?

Apply through the Anthropic careers site, or use this LinkedIn strategy to reach an Anthropic recruiter or engineer directly.

Anthropic is growing fast and gets a lot of applicants for every role. Pick the roles that actually match your background instead of applying to everything. Product, infrastructure, and research engineering roles all have different loops, so the job post tells you a lot about what you'll be tested on.

What topics should I study for?

• Dictionaries (almost every practical problem needs one)
• Heaps
• Intervals
• Graph (and tree) traversal
  • Graph breadth first search (web crawler questions are basically BFS)
  • Graph depth first search
• Sliding window
• Object-oriented design, like building a small class that grows new features over time
• Concurrency (threads, locks, and queues), especially for infrastructure roles
• System design
• Reading and parsing files, JSON, and logs in Python

Anthropic's coding rounds look more like real work than LeetCode. You'll usually build something that runs, then keep adding to it as the interviewer gives you new requirements. Python is the language most candidates report using. Get fast with the standard library, because you can look up docs but you can't use AI tools unless the instructions say you can.

What are some example Anthropic technical questions?

• Build an in-memory database that supports get, set, and delete, then add filtering, then add timestamps and expiration.
• Build a simple banking system with deposits, transfers, and account history.
• Build a task manager where tasks have priorities and can depend on other tasks.
• Write a web crawler that visits every page on one site, then make it run on multiple threads.
• Find all the duplicate files in a folder.
• Parse a pile of stack traces and report which functions show up the most.
• Scale and transform an image stored as a grid of numbers.
• Design a chat system.
• Design a system that sends very large model files to thousands of machines.
• Design a service that batches requests to a model so the GPUs stay busy.

The first three are the kind of thing that shows up on the CodeSignal test. They're split into levels, and you only unlock the next level once the tests for the current one pass. Candidates say you can move on without a perfect score, so don't burn twenty minutes on one failing edge case.

What are some example Anthropic behavioral questions?

• Why Anthropic? Why now?
• What do you think AI will do to society over the next few years? What worries you?
• Tell me about a time you changed your mind about something important.
• Tell me about a time you had to choose between what was good for you and what was right.
• Tell me about a project you owned from start to finish.
• Tell me about a disagreement with a teammate. How did it end?

Anthropic has a separate culture interview, and it counts as much as the technical ones. Candidates and prep sites say a weak culture interview can sink an otherwise strong loop. You don't need to be an AI safety researcher, but you should have actually thought about the risks and benefits of AI and be able to talk about them in your own words. Read a couple of Anthropic's own posts about their mission before the call.

Don't try to just agree with your interviewer. Interviewers want to see that you can hold an opinion, explain why, and change it when you hear a good reason. A real story about being wrong is worth more than a polished story about being right.

How long is the entire Anthropic interview process?

Usually three to six weeks from the recruiter call to an offer, sometimes longer. Anthropic checks references after the final round, sometimes while it figures out team match, and that step can add a few weeks of quiet. Line up two or three references before your onsite (a former manager and a close teammate are ideal) so you're not scrambling. And keep your other interviews going while you wait.

What is the Anthropic interview like for a full time software engineering role?

Recruiter call (15-30 minutes)

Your background, why you want to work on AI, and why Anthropic in particular. Have a real answer ready. This call counts more here than at most companies.

CodeSignal assessment (about 90 minutes)

A timed project, not a set of puzzles. You build one small system in stages, like the in-memory database above, and each stage adds new requirements on top of your code. Write clean code from the start, because you'll be changing it three more times before the end.

Live technical screen (about 1 hour)

A practical coding problem in a shared editor or notebook, like finding duplicate files or processing a batch of stack traces. The interviewer wants working code you can run, not pseudocode.

Virtual onsite

Typical composition:

• 1-2 more practical coding interviews. Infrastructure roles often get a concurrency problem, like making the web crawler multithreaded.
• 1 system design interview. Product roles focus on the API, data model, and user flow. Infrastructure roles go deeper on throughput, scheduling, and what happens when a machine dies.
• 1 hiring manager interview about projects you've owned, hard production problems, and how you make decisions without all the information
• 1 culture interview about the mission and how you think about AI

Reference checks and team match

These come after the onsite. Your process isn't over until references are done.

What are the levels and salary for a software engineer at Anthropic?

*Note that compensation varies based on location. These estimates are likely based on the USA HCOL market.

Anthropic doesn't use the long ladder of levels you'll see at Google or Meta, so the titles above are rough. A big share of the package is equity in a private company. That means you can't sell it whenever you want, and its value depends on future funding rounds and buybacks. Ask your recruiter how and when employees have been able to sell shares before you compare an offer to one with public stock.

How should I prepare?

Practice building small systems in Python from scratch, then adding features to them under a timer. A good drill is to write an in-memory key-value store in 20 minutes, then give yourself a new requirement every 15 minutes. Do a few concurrency problems if you're applying for infrastructure. For the culture interview, write down what you actually think about AI's risks and benefits, and one story where you changed your mind. Start with how to prepare for a technical interview if you're early in your prep.