Research Engineer, Knowledge Systems & RAG (Remote)

Anthropic Limited; United States; Hybrid; USD 350,000 - 850,000; Full time

Job summary

Anthropic is seeking an AI Engineer to advance research and production software for language models in a remote-friendly, US-based role with travel. You will design architectures, train models to interact with external data sources, and build robust evaluation methods to reveal system strengths and weaknesses.

You will collaborate with product and research teams to develop scalable retrieval, knowledge graphs, and agentic search capabilities, while delivering reliable code and aligning with

Qualifications

• Proven Python programming skills with reliable, maintainable code.
• Practical machine-learning research experience.
• Experience building software that uses large language models.
• Strong written and verbal communication with teamwork mindset.
• Willingness to contribute beyond a narrow job description.

Responsibilities

• Design and implement strategies for structuring information into knowledge bases.
• Fine-tune language models and apply reinforcement learning to improve interactions.
• Build evaluation datasets to expose failure modes in external-data use.
• Design and assess advanced agentic search capabilities.
• Collaborate with product and research teams to deliver capabilities.
• Develop scalable retrieval systems involving search, indexing, ranking, and knowledge graphs.

Skills

Python

Machine learning

Information retrieval

Communication

Pair programming

Job description

Anthropic is seeking an AI Engineer to advance research and production software for language models in a remote-friendly, US-based role with travel. You will design architectures, train models to interact with external data sources, and build robust evaluation methods to reveal system strengths and weaknesses.

You will collaborate with product and research teams to develop scalable retrieval, knowledge graphs, and agentic search capabilities, while delivering reliable code and aligning with