Research Engineer, Production Model Post-Training

Anthropic seeks a Senior Research Engineer in Zürich to lead post-training efforts for production AI models, combining research and engineering to improve model safety, alignment, and performance. The role involves scaling post-training techniques and collaborating across teams to deliver reliable, high-quality models.

As a Research Engineer on the Post-Training team, you will play a key role in advancing Anthropic’s production models by applying and refining post-training methods such as Constitutional AI and RLHF. Your work bridges research and engineering, directly shaping the safety, alignment, and capabilities of the Claude models deployed in production.

Responsibilities

• Develop and optimize post-training methods for large frontier models at scale
• Research and refine training recipes that enhance the quality of production models
• Create and maintain efficient, reliable pipelines for fine-tuning and evaluation
• Build tools to assess and improve model performance across multiple dimensions
• Work with research teams to adapt new techniques into production systems
• Investigate and resolve complex issues in training pipelines and model behavior
• Help define standards for reproducible and dependable model post-training

Requirements

• Strong skills in Python programming
• Experience training, fine-tuning, or evaluating large language models
• Solid software engineering background with experience in complex ML systems
• Familiarity with distributed computing and high-performance environments
• Ability to balance exploratory research with engineering precision and system reliability
• Proficiency in analyzing and troubleshooting model training processes
• Capability to work effectively in dynamic, evolving research settings

Nice to have

• Prior experience with large language models
• Strong interest in AI safety and ethical deployment practices

What the company offers

• Competitive salary and comprehensive benefits package
• Optional equity donation matching program
• Generous vacation and extended parental leave
• Flexible working hours to support work-life balance
• Collaborative office space in Zürich for team interaction

About the company

Anthropic aims to build AI systems that are safe, interpretable, and beneficial for users and society. The company approaches AI research as an empirical science, emphasizing large-scale, collaborative efforts to achieve impactful results. It fosters a culture of teamwork, diverse perspectives, and frequent research discussions to ensure progress toward steerable, trustworthy AI.

• Focus on creating reliable, steerable AI with broad societal benefit
• Commitment to large-scale, collaborative research efforts
• Emphasis on empirical science and high-impact work over narrow problems
• Regular research discussions to align on most valuable projects
• Value diversity of perspectives in team composition