Latest AI JobsBrowse Jobs
Anthropic logo

Research Engineer, Pretraining Scaling

AnthropicSan Francisco, CAAI Research & Engineering
Apply on company site
Salary
Not listed
Job type
Full-time
Posted
30 Sept 2025
Status
Open · checked 7d ago

About this role

The Research Engineer, Pretraining Scaling position at Anthropic is an in-office role based in San Francisco, California, focusing on production pretraining pipelines and large-scale machine learning systems. This position suits experienced professionals who possess a strong blend of research curiosity and engineering rigor to build safe and beneficial AI systems.

Role Overview

Anthropic is hiring a Research Engineer for its ML Performance and Scaling team in San Francisco, CA. The role operates at the intersection of research and engineering, requiring hands-on work across the entire production training stack. Team members optimize performance, debug hardware issues, design experiments, and coordinate launches for frontier models. This position demands operational agility, responsiveness during high-stakes launches, and a commitment to creating reliable and interpretable AI systems.

Key Responsibilities

Own core components of the production pretraining pipeline, focusing on model operations, reliability, observability, and performance optimization.
Troubleshoot and resolve intricate challenges spanning the full stack, covering networking, hardware errors, evaluation infrastructure, and training dynamics.
Design and execute experiments aimed at increasing uptime, cutting step time, improving training efficiency, and boosting overall model performance.
Manage on-call duties during model launches by swiftly diagnosing production roadblocks and coordinating resolutions across cross-functional teams.
Develop and maintain evaluation infrastructure, production logging, and monitoring dashboards.
Integrate new functionalities into the training codebase, such as novel architectures or long context support.
Coordinate closely with colleagues across San Francisco and London alongside Architectures, Tokens, and Systems teams.
Document debugging procedures, lessons learned, and system architectures to expand institutional knowledge.

Requirements

Hands-on background in training large language models or deep familiarity with large-scale distributed systems, JAX, TPU, or PyTorch.
Enthusiasm for balancing research and engineering tasks evenly.
Willingness to participate in on-call rotations for production systems, manage extended hours during launches, and tackle difficult challenges under tight pressure.
Ability to pivot priorities flexibly based on the immediate needs of the production model.
Strong troubleshooting abilities for ambiguous and complex problems across multiple layers of the technology stack.
Clear communication skills and collaborative aptitude for working across different time zones and high-stress scenarios.
Passion for refining craft as a research engineer and addressing the societal impacts of artificial intelligence.

Nice to Have

Prior experience training large language models or utilizing ML frameworks at scale with JAX, TPU, or PyTorch.
Contributions to open-source large language model frameworks such as open_lm, mesh-transformer-jax, or llm-foundry.
Published academic or industry research focused on scaling laws, model training, or machine learning systems.
Background working with production machine learning systems, evaluation infrastructure, or observability tools.
Previous professional experience as a quantitative specialist, systems engineer, or in roles demanding operational excellence and technical depth.

Skills and Tools

PyTorch
LLM
JAX
TPU
Distributed Systems
Production ML Frameworks

Salary and Benefits

The annual compensation range for this role is $350,000 to $850,000 USD. Anthropic provides competitive compensation, optional equity donation matching, flexible working hours, generous parental leave and vacation policies, and a collaborative office environment.

About the Company

Anthropic is a public benefit corporation headquartered in San Francisco focused on building reliable, steerable, and interpretable artificial intelligence systems. Operating as a single cohesive team tackling large-scale empirical science research, the company brings together researchers, engineers, policy experts, and business leaders dedicated to safe and beneficial AI.

How to Apply

Interested candidates should visit the official Anthropic careers page to submit an application for this position.

Skills and tech stack

PyTorchLLM

Questions candidates ask

Where is this job located?

This position is located in San Francisco, CA, and requires working in-office 5 days per week.

What is the salary range for this role?

The annual compensation range is between $350,000 and $850,000 USD.

What skills and frameworks are relevant for this position?

Relevant technologies include PyTorch, JAX, TPUs, large language models, large-scale distributed systems, and production ML frameworks.

Does Anthropic offer visa sponsorship?

Yes, Anthropic sponsors visas and retains an immigration lawyer to assist candidates who receive an offer.

How can I apply?

Candidates can apply directly through the official Anthropic careers page.

Ready to apply?

The application is on the official Anthropic careers page.

Apply on company site