AI Security Engineer - Offensive & Blue-Team

Metr; Berkeley (CA); Hybrid; USD 180,000 - 260,000; Full time

Benefits offered by this job

Hybrid work in Berkeley

Visa sponsorship (cap-exempt H-1B)

Diversity and equal opportunity

Job summary

METR is actively building a security-focused team in Berkeley to safeguard frontier AI research and evaluation infrastructure. The role emphasizes offensive security, threat detection at scale, and secure cloud environments across AWS and container platforms.

You will work with researchers to ensure safe pre-deployment evaluations, respond to evolving threats, and help protect confidential materials while enabling cutting-edge AI safety studies in a hybrid in-office setting.

Qualifications

• Experience securing AI systems and cloud environments.
• Proven background in red-teaming, threat modeling, or adversarial research.
• Familiarity with modern SIEM/detection pipelines and incident response.

Responsibilities

• Conduct targeted red-team exercises against internal systems and build automated AI red teaming.
• Develop high-context detection and rapid triage capabilities for threats from internal and external actors.
• Lead blue-team engineering: telemetry pipelines, incident response, and cloud/infrastructure hardening.
• Secure a unique attack surface by evaluating frontier AI agents and pre-deployment evaluations.
• Collaborate with researchers to make dangerous-capability experiments safe to run and defend against internal threats.

Skills

Deep security

Offensive security

AI/LLM engineering

AWS knowledge

Tools

AWS

EKS

Lambda

ECS

IAM

SQS

CloudWatch

SecurityHub

GuardDuty

PostgreSQL

Pulumi

DataDog

Okta

Google Workspace

Tailscale

CrowdStrike Falcon

Kubernetes

Job description

METR is actively building a security-focused team in Berkeley to safeguard frontier AI research and evaluation infrastructure. The role emphasizes offensive security, threat detection at scale, and secure cloud environments across AWS and container platforms.

You will work with researchers to ensure safe pre-deployment evaluations, respond to evolving threats, and help protect confidential materials while enabling cutting-edge AI safety studies in a hybrid in-office setting.