Research Engineer, Pretraining Scaling

Overview

Research Engineer on Anthropic’s ML Performance and Scaling team, focused on ensuring production pretrained frontier models train reliably, efficiently, and at scale.

What you'll do

• Own critical production aspects of the pretraining pipeline, including model operations, performance optimization, observability, and reliability.
• Debug and resolve complex full-stack issues spanning hardware errors, networking, training dynamics, and evaluation infrastructure.
• Design and run experiments to improve training efficiency, reduce step time, increase uptime, and enhance model performance.
• Respond to on-call incidents during model launches and coordinate solutions across teams.
• Build and maintain production logging, monitoring dashboards, and evaluation infrastructure.
• Add capabilities to the training codebase, such as long context support or novel architectures.

What you'll need

• Hands-on experience training large language models, or deep expertise with JAX, TPU, PyTorch, or large-scale distributed systems.
• Experience and interest in both research and engineering work (described as roughly 50/50).
• Excitement about being on-call for production systems, including long days during launches.
• Ability to debug complex, ambiguous problems across multiple layers of the stack.
• Clear communication and effective collaboration, including when coordinating across time zones.
• Passion for refining craft as a research engineer.
• Care about societal impacts of AI and responsible scaling.

Details

• Location: in-office 5 days per week in San Francisco.
• Office time policy: staff expected to be in an office at least 25% of the time, with some roles requiring more.
• Visa sponsorship: sponsored for some roles/candidates; not guaranteed.