jobs

Research Engineer, Pretraining Scaling

Anthropic’s ML Performance and Scaling team trains production pretrained models that directly shape the company’s future AI systems. This Research Engineer role focuses on ensuring frontier models train reliably, efficiently, and at scale.

Company

Anthropic

Location

San Francisco, United States

Status

Closed

Posted

2026-09-13T03:31:18.092+00:00

Anthropic's ML Performance and Scaling team is hiring a research engineer to keep production pretrained models training reliably and efficiently at frontier scale. The work spans performance optimization, hardware and networking debugging, experimental design, observability, and launch coordination, plus on-call incidents during model launches. The posting asks for hands-on large language model training experience or deep expertise in JAX, TPU, PyTorch, or large-scale distributed systems, and comfort with operational responsibility. It is a full-time role based in the San Francisco office five days a week, with pay listed at 350,000–850,000 USD.