Research Scientist, Tokens (Multimodal)

Details

Anthropic’s Multimodal team builds and studies multimodal models (images, video, audio, and text) to better understand and mitigate the risks introduced by powerful multimodal AIs. The team works across training, inference, system design and data collection, and focuses on foundational research, infrastructure, and large-scale data ingestion and tooling. The role is appropriate for candidates along the researcher/engineer curve — from strong engineers to experienced researchers — who want to work on large-scale empirical AI research.

Responsibilities

• Develop new architectures for modeling multimodal data and study interactions with text-only models at scale.
• Build and maintain infrastructure for multimodal systems, including high-performance RPC servers for image inputs and sandboxing infrastructure for secure data collection.
• Design and implement complex multimodal reinforcement learning environments and other experimental platforms.
• Develop tooling to collect, process, and clean very large multimodal datasets (large-scale ETL and data ingestion pipelines).
• Run simple experiments at very large scale rather than smaller complex experiments; iterate and scale infrastructure and experiments accordingly.
• Collaborate closely with researchers and engineers (pair programming is encouraged) and contribute across the stack from training to inference and data collection.

Requirements

• At least a Bachelor's degree in a related field or equivalent experience.
• Significant software engineering experience; results-oriented with a bias toward flexibility and impact.
• Interest in and willingness to learn more about machine learning research and the societal impacts of AI systems.
• Strong candidates may also have experience with:
  • High-performance, large-scale ML systems
  • GPUs, Kubernetes, PyTorch, or OS internals
  • Language modeling with transformers
  • Reinforcement learning
  • Large-scale ETL and data ingestion tooling
• Comfortable working collaboratively (pair programming) and across role boundaries when needed.

Logistics

• Locations: San Francisco, CA; New York City, NY; Seattle, WA (United States).
• Location-based hybrid policy: staff are expected to be in one of the offices at least 25% of the time; some roles may require more office time.
• Visa sponsorship: Anthropic does sponsor visas and retains immigration counsel, though sponsorship may not be possible for every role or candidate.

Benefits

• Competitive compensation (see salary range).
• Optional equity donation matching, generous vacation and parental leave, flexible working hours, and collaborative office spaces.
• Emphasis on research impact, frequent research discussions, and a collaborative environment.

How We Work / Culture

• Work is organized as large-scale research efforts with an emphasis on impact and empirical science.
• Frequent research discussions and collaboration; strong communication skills are valued.
• Encouragement for applicants from diverse backgrounds; Anthropic values representation and inclusion in its teams.