You and the process

With a person
Anna, in-house recruiter
9 years hiring engineers
30 minutes with a real recruiter
They read your CV with you, on a call, and say where the offers are being lost.
Didn’t find what you were looking for? Tell us what to build
Anthropic

Research Engineer, AI Monitoring

  • Hybrid

Salary

£375,000 - 640,000/ year

AI summary

For members

The whole posting in a few lines. Sign up to read it here and on every role you open.

Sign Up to Read

Description

The team builds, tests, deploys and monitors classifiers that stop internal agents from taking dangerous actions, such as sandbox escapes or exfiltrating sensitive data.

The classifiers run across RL, evaluations and internal deployments. When one flags an action, it can block the action before the tool call executes, end the task and alert a human.

This is one layer in a defence-in-depth setup. Flagged actions feed offline monitoring, and those findings improve the classifiers and the design of safer environments.

The work mixes research and production engineering. The research side means building classifiers with high recall and high precision for rare, novel and sometimes adversarial behaviour, plus narrow questions such as deploying safety classifiers inside training without pushing models to evade them. Some of this builds on the multi-stage designs described in the team's work on Claude Code auto mode and Constitutional Classifiers.

The engineering side means putting these classifiers on the critical path of research and product work, covering very high traffic volumes within cost, latency and availability limits.

Much of the daily work is done with Claude: you direct agents to run experiments and ship code, then review their output. The hard part becomes deciding what to do and in what order. Organisation, communication and prioritisation matter a lot, since the classifiers sit in the path of many other teams' work.

The company is a public benefit corporation headquartered in San Francisco. Its stated mission is to build reliable, interpretable and steerable AI systems. The team is growing quickly and includes researchers, engineers, policy experts and business leaders.

Responsibilities

  • Design, train and evaluate classifiers that detect unsafe agent actions in real time, which can include prompted models, trained classifiers and probes on model internals.
  • Deploy, tune and run these classifiers on the critical path of internal agents, RL training and evaluations, and own their accuracy, reliability, latency, cost and failure behaviour.
  • Build the datasets, labelling pipelines and evaluations used to measure and improve the classifiers and harm taxonomies.
  • Work with red teaming colleagues to stress-test the classifiers and strengthen them based on what they find.
  • Work with the teams whose agents, environments and evaluations the classifiers run on, so false positives are fixed quickly without weakening protection.
  • Help set the team's direction, including what to build, what to partner on and where to invest.

Requirements

  • You can organise and prioritise several parallel streams of work and communicate clearly with the teams affected.
  • You can critically review code and experiment results, including output produced by AI agents.
  • You have a sincere interest in developing safe and beneficial AI systems.
  • Preferred: experience leading ambiguous technical projects and setting direction for other engineers or researchers.
  • Preferred: experience evaluating large language models, building classifiers, or running high-scale production systems where requirements change quickly.
  • Preferred: experience building detection systems where true positives are rare and false positives are costly, such as abuse, fraud, spam or intrusion detection.
  • Preferred: experience running low-latency, high-availability services that other systems depend on, including on-call duty for them.
  • Preferred: experience fine-tuning language models, or working with probes and other methods that read model internals.
  • Preferred: experience with LLM inference or serving infrastructure, or integrating new components into training infrastructure.
  • Preferred: background in security, such as sandboxing or detection and response.
  • Preferred: familiarity with research on AI control, scalable oversight, agent safety or reinforcement learning.
  • Minimum education is a Bachelor's degree or an equivalent mix of education, training and/or experience, in a field relevant to the role as shown through coursework, training or professional experience.
  • Required years of experience depend on the internal job level for the position.
  • You are encouraged to apply even if you do not meet every qualification. The company wants a range of perspectives on the team.

Conditions

Annual salary range: £375,000 to £640,000 GBP.

Location-based hybrid policy: staff are currently expected to be in an office at least 25% of the time, and some roles may require more.

Visa sponsorship is available, but not for every role or every candidate. If an offer is made, the company will make every reasonable effort to obtain a visa and uses an immigration lawyer to help.

Recruiters contact candidates only from company's site email addresses, and may occasionally use vetted agencies acting on the company's behalf. Recruiters never ask for money, fees or banking details before your first day.

Benefits

  • Competitive compensation and benefits.
  • Optional equity donation matching.
  • Generous vacation and parental leave.
  • Flexible working hours.
  • An office space for collaborating with colleagues.

Where you’d work

Part of the week in the office

You can work from

  • United Kingdom

Visa sponsored

About the company

Anthropic

  • Industry: AI

Office in London, United Kingdom

Also hiring in San Francisco, United States, New York City, United States, Seattle, United States and 7 more places

11 of their 119 open roles are remote

Your chances

Worth a look before you spend an evening tailoring a CV for it.

  • 22 checks run
  • 3 red flags

Still hiring?

14 checks

No red flags

How crowded?

8 checks

3 red flags

Fits Me

How well does this role fit you?

Answer a few questions or drop your CV, and every role gets a fit score with the reasons, this one first.

  • Your field
  • Level
  • Stack
  • Work model
  • Salary floor
  • Must-haves
Details12 facts · Role, Location, Compensation, Company
Tech stack
  • LLMs
Equity
Equity offered
Industry
AI
Specialty
ML
Region
United Kingdom
Pay period
Annual
Show 6 more factsShow less

Role

Category
Data & Analytics
Specialty
ML
Tech stack
  • LLMs

Location

Work model
Hybrid
Region
United Kingdom
Office
  • London, United Kingdom
Remote from
  • United Kingdom
Visa sponsorship
Sponsored

Compensation

Salary
£375,000 - 640,000 / year
Pay period
Annual
Equity
Equity offered

Company

Industry
AI

Something wrong with this vacancy?

Similar vacancies

Share this vacancy

What's wrong with it?

The employer never sees who reported.

Reason