SRE Lead, Safeguards ML Infra & Safe Deploys

Fdetimes; San Francisco (CA); Ibrido; USD 320.000 - 485.000; Tempo pieno

Descrizione del lavoro

Anthropic is seeking an engineer to own and automate Safeguards deployment for model launches, ensuring robust safety systems across platforms (1P, AWS Bedrock, GCP Vertex). You will lead canary rollouts, configure safeguards, and drive continuous validation to reduce manual work.

You will bring hands-on experience with production releases, incident response, and drive efficiency by turning runbooks into tooling, while collaborating across research and platform teams in a hybrid,

Competenze

• Experience in production change management at scale and deployment pipelines.
• Experience conducting launch readiness or incident response in production environments.
• Ability to automate deployment and validation tasks into repeatable pipelines.

Mansioni

• Launch captain model releases: stand up, configure, and verify safeguards for every new model.
• Own off-cycle deployment of new safety classifiers with canarying and post-deploy validation.
• Verify safeguards are live across all deployment platforms and detect drift.
• Automate hand-built checks into continuous validation and pipelines.
• Turn runbooks into tooling and one-off deploys into repeatable processes.
• Participate in on-call rotations covering incidents and time-sensitive launches.

Conoscenze

Launch captain

Incident commander

Release owner

Automation

Production change management

Cloud platforms (AWS, GCP)

Python

System reliability

Formazione

Bachelor’s degree or equivalent

Strumenti

AWS

GCP

Rust

Python

Descrizione del lavoro

Anthropic is seeking an engineer to own and automate Safeguards deployment for model launches, ensuring robust safety systems across platforms (1P, AWS Bedrock, GCP Vertex). You will lead canary rollouts, configure safeguards, and drive continuous validation to reduce manual work.

You will bring hands-on experience with production releases, incident response, and drive efficiency by turning runbooks into tooling, while collaborating across research and platform teams in a hybrid,