Research Engineer, AI Monitoring

Anthropic PBC
  • Full-time
  • San Francisco, United States
Development Research

Anthropic is seeking an engineer-researcher to build and operate classifiers that detect and block unsafe actions by internal AI agents. The role combines research into accurate detection of rare, novel, and adversarial behavior with production engineering for systems used in research, training, and evaluations. You will help make these safeguards reliable at scale while working closely with other teams and using AI agents to run experiments and ship code.

Requirements

  • Organize and prioritize multiple workstreams, and communicate clearly with affected teams.
  • Critically review code and experimental results, including work produced by AI agents.
  • Have a genuine interest in developing safe and beneficial AI systems.
  • Design, train, and evaluate real-time classifiers for unsafe agent actions, such as prompted models, trained classifiers, or probes into model internals.
  • Deploy, tune, and operate classifiers in internal agents, reinforcement-learning training, and evaluations; take responsibility for accuracy, reliability, latency, cost, and failure behavior.
  • Build datasets, labeling pipelines, and evaluations to measure and improve classifiers and harm taxonomies.
  • Partner with red-team colleagues to stress-test classifiers and strengthen them against discovered weaknesses; work with teams using the systems to resolve false positives without weakening protections.
  • Contribute to team direction, including decisions about what to build, where to partner, and where to invest.
  • Preferred experience includes leading ambiguous technical projects; evaluating large language models, developing classifiers, or running high-scale production systems; and building detection systems where true positives are rare and false positives costly, such as abuse, fraud, spam, or intrusion detection.
  • Further useful experience includes operating low-latency, highly available services and being on call; fine-tuning language models or using probes into model internals; LLM inference or serving infrastructure; integrating components into training infrastructure; security; or research in AI control, scalable oversight, agent safety, or reinforcement learning.
  • A bachelor’s degree or equivalent education, training, and/or experience is the minimum stated education requirement; relevant study or experience is required. Experience expectations depend on the internal job level.

Additional information

Anthropic is a public benefit corporation headquartered in San Francisco, focused on reliable, interpretable, and steerable AI. The role’s annual salary range is £375,000–£640,000 GBP. Staff are currently expected to work from an office at least 25% of the time, with some roles requiring more. Visa sponsorship may be available, and the company encourages interested candidates to apply even if they do not meet every qualification. Benefits include competitive compensation, optional equity donation matching, generous vacation and parental leave, flexible hours, and collaborative office space. Applicants can review Anthropic’s guidance on AI use in applications; recruiters contact candidates from @anthropic.com or identify vetted recruiting agencies, and will never request money or banking information.

More jobs within the Development and Research categories

APG Asset Management

  • Full-time
  • Heerlen, Netherlands

Salary: 5533.19 - 7143.86 EUR P1M

Development Research

Companies hiring AI talent

From research labs to fast-growing startups, these employers are building with Artificial Intelligence and recruiting on OnlyAIJobs.