Anthropic

Official guidance

Anthropic develops the Claude family of AI models and the products built on them, including the Claude apps, Claude Code and an API platform. Its careers page says engineers do substantial research and researchers do substantial engineering. Work spans model training and evaluation, safety and interpretability research, training and inference infrastructure, and product and platform engineering.

Relevant roles: General software engineering, Infrastructure / platform engineering, ML engineering, Research-oriented ML implementation, Training / inference infrastructure, Research engineering. Domains: Algorithms, Software engineering, ML & applied AI, AI infrastructure, AI research, Interview skills.

What the sources establish

Anthropic publishes a short official description of its interviews (all over Google Meet; technical roles use live coding tools such as Colab and CodeSignal, where candidates may look things up) and a dated AI-usage policy (last updated July 2025): AI is encouraged for refining applications and preparing, but take-home assessments and live interviews are without AI unless Anthropic explicitly allows it. Not established: the number and sequence of interviews, the topics of technical rounds, languages, take-home formats and level differences.

Careers at Anthropic (How we hire section)
Official
Anthropic. Undated page; read 2026-10-05. Read in full for this page.
Scope: All Anthropic roles, with separate notes for technical and non-technical roles
Documented by this source
- All interviews are conducted over Google Meet
- Technical roles use live coding tools such as Colab and CodeSignal; candidates may look things up but should be comfortable with basic syntax and standard libraries
- Interviews also cover your experience and motivation, with time for your own questions
- States that about half of the technical staff had no prior ML experience; independent research, thoughtful blog posts and open-source contributions are worth putting at the top of a résumé
- Engineers do substantial research and researchers do substantial engineering; candidates with an engineering background are advised to apply as engineers
- No internships are currently offered; no feedback is given on applications or interviews; reapplication is possible after 12 months, or sooner after a material change
Outside or AI tools, according to this source: For technical interviews you can look things up but should know basic syntax and standard libraries; AI use is governed by the separate candidate AI guidance. Follow the instructions in your own assessment invitation.
- Undated; does not give the number, length or order of interviews or the topics of technical rounds.
- It does not define what looking things up covers; it does not override the AI guidance, which excludes AI assistance in live interviews unless stated.
Open source(opens in a new tab)Link responded, 5 Oct 2026
Guidance on Candidates’ AI Usage
Official
Anthropic. Last updated 2025-07-10 (per the page); read 2026-10-05. Read in full for this page.
Scope: All Anthropic applicants, covering applications, take-home assessments, interview preparation and live interviews
Documented by this source
- Applications: write your own first draft, then you may use Claude to refine it; using AI to invent experiences is not allowed
- Take-home assessments: complete them without Claude unless Anthropic explicitly says AI is allowed for that assessment
- Interview preparation: using Claude to research Anthropic, practice answers and prepare questions is encouraged
- Live interviews: no AI assistance unless Anthropic indicates otherwise; accommodation requests should go to the recruiter early
- Describes how Anthropic itself uses Claude in hiring (job descriptions, interview questions, communications, metrics, transcription, sourcing) and states that Claude does not make hiring decisions
Outside or AI tools, according to this source: AI is encouraged for refining applications and for preparation; take-home assessments and live interviews are done without AI unless Anthropic explicitly allows it for that step. Follow the instructions in your own assessment invitation.
- The page says the guidance will be reviewed and updated regularly; check the current version before your interviews.
Open source(opens in a new tab)Link responded, 5 Oct 2026

Our preparation recommendation

This is our suggestion based on the role and the sources above. It is not a statement by Anthropic.

Our recommendation, not Anthropic guidance: the documented environment (live coding in tools such as Colab or CodeSignal, lookups allowed, no AI unless stated) rewards fluent, practical coding in a language whose syntax and standard library you know well. Colab is a Python notebook environment, so Python fluency is a sensible default unless your recruiter says otherwise. Research and ML roles should add model and framework depth, and everyone should prepare clear accounts of independent work, since the careers page highlights research, writing and open source.

Coding fluency and standard-library costsNot started
Lookups are allowed, but basic syntax and standard libraries should not cost you time.
Python fundamentals for interviews and research codeNot started
Colab is a Python notebook environment; confirm the language for your interview (our inference).
Pattern recognition and mixed practiceNot started
Live coding in tools such as CodeSignal benefits from broad, practiced problem patterns.
Debugging, code reading, review and refactoringNot started
Live coding in tools that run code means iterating and debugging under time pressure.
Transformers, attention and language-model objectivesNot started
For ML and research-engineering roles working on large language models.
PyTorch and autograd fluencyNot started
For research and ML roles: framework fluency for notebook-based implementation work.
Project deep divesNot started
The careers page highlights independent research, writing and open source; be ready to discuss yours in depth.
Behavioral stories and values-based questionsNot started
Interviews ask about your experience and motivation.
Assessment formats, tool policies and logisticsNot started
AI is off-limits in take-homes and live interviews unless stated; know the rules for each step.

Confirm with your recruiter

The full sequence of interviews for your role and level: screens, take-homes, technical rounds and their lengths.
Which live coding environment you will use (Colab, CodeSignal or another) and which language is expected.
What looking things up covers in your technical interviews (documentation, web search) as opposed to AI assistants, which the AI guidance excludes unless stated.
For any take-home: time limit, deadline, allowed resources, and whether AI use is explicitly permitted for that assessment.
Whether ML depth (training, evaluation, model internals) is assessed for your role.
Logistics: Google Meet setup, time zones and accommodations.

Limitations of this page

The pages read describe the interview environment and AI rules, not the content of technical rounds.
Research, engineering and non-technical loops are described only in a sentence or two each.