Research Scientist, Tokens (Multimodal)

Details

Anthropic’s mission is to create reliable, interpretable, and steerable AI systems that are safe and beneficial for users and society. The Multimodal team focuses on building and studying models that operate on multiple data modes (text, images, video, audio) to augment human creativity while mitigating risks associated with powerful multimodal AIs.

Responsibilities

• Develop new architectures for modeling multimodal data and study their interaction with text-only models at scale.
• Build infrastructure including complex multimodal reinforcement learning environments, high-performance RPC servers for image processing, and secure data collection sandboxing.
• Develop tooling for collecting, processing, and cleaning large-scale multimodal data to support large-scale experiments.
• Collaborate as part of a multidisciplinary team, with a bias towards flexibility and impact, and engage in pair programming.

Requirements

• Significant software engineering experience.
• Results-oriented with flexibility and a willingness to take on tasks beyond the job description.
• Interest in learning more about machine learning research and care about societal impacts of AI.
• Strong candidates may also have experience with large-scale ML systems, GPUs, Kubernetes, PyTorch, OS internals, language modeling with transformers, reinforcement learning, and large-scale ETL.
• Bachelor's degree in a related field or equivalent experience is required.

Benefits

• Competitive compensation with an annual salary range of $280,000 - $425,000 USD.
• Visa sponsorship where possible.
• Hybrid work policy requiring at least 25% in-office presence.
• Optional equity donation matching, generous vacation and parental leave, flexible working hours, and collaborative office spaces.
• Strong emphasis on diversity, inclusion, and the societal and ethical implications of AI research.

How We're Different

• Work on large-scale, impactful AI research as a cohesive team.
• Emphasis on empirical science and collaboration.
• Frequent research discussions to prioritize high-impact work.
• Research builds on successful prior work including GPT-3, AI safety, and interpretability.