Research Engineer, AI Monitoring
£375k - £640k • London, UK • Full-time
Posted 4d ago
Job Location
London, UK
Tech Stack
Remote Work Policy
On-site
Employment Type
Full-time
Visa Sponsorship
The posting says they sponsor visas
Categories
AI Engineer
About the job
Anthropic is building reliable, interpretable, and steerable AI systems to ensure AI is safe and beneficial for users and society. This role focuses on developing, testing, deploying, and monitoring classifiers that prevent internal AI agents from taking dangerous actions, such as attempting to escape sandboxes or exfiltrate sensitive data. These classifiers are a critical layer in a defense-in-depth strategy, feeding into offline monitoring and informing improvements to classifiers and environment design. The work involves both research into high-recall, high-precision classifiers for rare and adversarial behaviors, and production engineering to ship these classifiers into the critical path of research and product workflows, ensuring they operate within cost, latency, and availability constraints. Day-to-day tasks involve directing AI agents to run experiments, shipping code, and reviewing their output, with a strong emphasis on organizational, communication, and prioritization skills.
Responsibilities
- Design, train, and evaluate classifiers for detecting unsafe agent actions in real-time, including prompted models, trained classifiers, and probes on model internals.
- Deploy, tune, and operate classifiers in the critical path of internal agents, RL training, and evaluations, ensuring accuracy, reliability, latency, cost, and failure behavior.
- Build datasets, labeling pipelines, and evaluations to measure and improve classifiers and harm taxonomies.
- Partner with red teaming colleagues to stress-test classifiers and harden them against identified vulnerabilities.
- Collaborate with teams whose agents, environments, and evaluations are impacted by classifiers to ensure quick resolution of false positives without weakening protection.
- Contribute to team direction, including product development, partnerships, and investment strategies.
Requirements
- Ability to organize and prioritize multiple parallel work streams.
- Clear communication skills with affected teams.
- Ability to critically review code and experiment results, including AI agent outputs.
- Sincere interest in developing safe and beneficial AI systems.
- Experience leading ambiguous technical projects and setting direction for others (preferred).
- Experience evaluating LLMs, developing classifiers, or operating high-scale production systems with rapidly changing requirements (preferred).
- Experience building detection systems for rare true positives and costly false positives (e.g., abuse, fraud, spam, intrusion detection) (preferred).
- Experience running low-latency, high-availability services with on-call responsibilities (preferred).
- Experience fine-tuning language models or working with probes and methods to read model internals (preferred).
- Experience with LLM inference/serving infrastructure or integrating 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 (preferred).
- Bachelor's degree or equivalent combination of education, training, and experience.
Benefits
- Annual Salary: £375,000—£640,000 GBP
- Visa sponsorship available
- Hybrid work policy (at least 25% in office)
