ML Engineer — Constitutional AI

Best AI Tools Wiki; San Francisco (CA); Hybrid; USD 180,000 - 300,000; Full time

Benefits offered by this job

Significant equity

Annual education stipend

Sabbatical after 4 years

Parental leave

Job summary

Anthropic is hiring an ML Engineer to build and scale Constitutional AI training systems and evaluation frameworks for the Claude family. You will implement novel training methodologies and optimize feedback loops that improve safety and usefulness.

This role demands strong Python and PyTorch/JAX skills, experience with RLHF or related preference learning, and solid software engineering practices. You will work in SF, contributing to cutting-edge AI safety research and scalable systems.

Qualifications

• 4+ years of ML engineering experience
• Proficiency in Python and PyTorch/JAX
• Experience with RLHF, DPO, or preference learning
• Understanding of language model training pipelines
• Strong software engineering practices

Responsibilities

• Build and scale Constitutional AI training systems
• Implement novel training methodologies
• Develop evaluation frameworks for AI models
• Optimize feedback loops to improve safety and helpfulness of Claude

Job description

Anthropic is hiring an ML Engineer to build and scale Constitutional AI training systems. You will implement novel training methodologies, develop evaluation frameworks, and optimize the feedback loops that make Claude safe and helpful.

Requirements

• 4+ years of ML engineering experience
• Strong proficiency in Python and PyTorch/JAX
• Experience with RLHF, DPO, or preference learning
• Understanding of language model training pipelines
• Strong software engineering practices

Nice to Have

• Experience with Constitutional AI or similar frameworks
• Background in AI safety or alignment research
• Experience with distributed training at scale
• Publications in relevant ML areas

Significant equity in a leading AI safety company

• Annual education stipend
• Sabbatical after 4 years
• Parental leave

Skills

Python PyTorch JAX RLHF Constitutional AI Distributed Training