Research Engineer, Pretraining Scaling – London

Overview

Join Anthropic’s ML Performance and Scaling team to keep production pretrained models training reliably at scale. You’ll operate at the research-engineering boundary, improving efficiency, observability, and reliability across the full training stack. Expect on-call incident response during launches and collaboration with global teams to push safe, steerable AI forward. This role offers hands-on work on large-scale ML systems with meaningful impact on our mission.

Pay / Benefits

• competitive compensation
• equity donation matching (optional)
• generous vacation and parental leave
• flexible working hours
• office space for collaboration
• visa sponsorship available

Responsibilities

• Own production pretraining pipeline components (model operations, performance optimization, observability, reliability)
• Debug complex issues across the stack—from hardware to training dynamics and evaluation infrastructure
• Design and run experiments to improve training efficiency and model performance
• Respond to on-call incidents during launches and coordinate cross-team solutions
• Build/maintain production logging, monitoring dashboards, and evaluation infrastructure
• Add capabilities to training codebase (e.g., long context support, new architectures)
• Collaborate with SF and London teams plus Tokens, Architectures, and Systems groups
• Document systems, debugging approaches, and lessons learned

Key requirements

• Hands-on experience training large language models or deep expertise with JAX, TPU, PyTorch, or large-scale distributed systems
• Enjoys both research and engineering work (ideally ~50/50 split)
• Willing to be on-call for production systems during launches
• Thrive in high-impact, dynamic environments and solving hard problems under pressure
• Strong communication and cross-time-zone collaboration skills
• Passion for AI safety and responsible scaling

• communication
• collaboration
• problem-solving under pressure
• JAX
• TPU
• PyTorch

…