Staff ML Engineer — Account Abuse & Fraud Detection

anthropic; San Francisco, New York (CA, NY); On-site; USD 320,000 - 485,000; Full time

Job summary

Anthropic is seeking a Staff+ Software Engineer for Account Abuse (Machine Learning) to build and operate ML systems that detect and stop abuse at scale in our SF offices. You will work across model training, productionization, and evaluation, using Claude to accelerate feature development and model deployment.

The role emphasizes production-ready ML, measurement, precision, and safe rollout, with collaboration across data scientists, policy, and platform teams to minimize latency impact and

Qualifications

• Bachelor’s degree or equivalent required.
• Experience with ML in production and feature platforms.
• Proficiency in Python and SQL required.

Responsibilities

• Build and operate a feature computation platform for model training and real-time scoring.
• Train, evaluate, and deploy models that detect account abuse and fraud.
• Build tooling to speed up feature development, training, and evaluation using Claude.
• Enable backtesting, shadow deployment, and safe rollout with monitoring for skew and drift.
• Collaborate with data scientists and Policy & Enforcement to improve label coverage and quality.
• Partner with product and platform teams to minimize latency impact while integrating model decisions.
• Strong communication skills to explain tradeoffs to non-technical stakeholders.

Skills

Python

SQL

ML deployment

Model training

Data pipelines

Communication

Education

Bachelor’s degree

Tools

Spark

Beam

Airflow

Chronon

Feast

Tecton

Flink

Kafka Streams

Job description

Anthropic is seeking a Staff+ Software Engineer for Account Abuse (Machine Learning) to build and operate ML systems that detect and stop abuse at scale in our SF offices. You will work across model training, productionization, and evaluation, using Claude to accelerate feature development and model deployment.

The role emphasizes production-ready ML, measurement, precision, and safe rollout, with collaboration across data scientists, policy, and platform teams to minimize latency impact and