- Anthropic opens two four-month fellowships in STEM and conceptual reasoning, offering US$100,000–US$167,000 in compensation. Applications close October 28, 2026, for programmes beginning January 12, 2027.
Artificial intelligence research company Anthropic invites applications for two four-month fellowship programmes designed to improve the reasoning capabilities of its AI models. The opportunities span science and engineering, as well as philosophy, cognitive science, economics, public policy and law. Applications close on October 28, 2026, with selected candidates expected to begin their fellowships in January 2027.
Both programmes run from January 12 to May 8, 2027, and offer compensation ranging from US$100,000 to US$167,000 for the four-month period, along with additional benefits. Fellows will work with Anthropic researchers to evaluate the capabilities and limitations of Claude, the company’s AI models, and investigate how AI can contribute to complex research problems.
The Conceptual Reasoning Fellowship targets researchers and experts in philosophy, cognitive science, economics, public policy and law. Participants will work with Anthropic’s research teams to assess Claude’s reasoning abilities in their respective disciplines and develop evaluations that test its performance on complex questions requiring expert judgement.
The work includes identifying important research challenges, developing benchmarks and datasets, and building testing environments that measure how effectively AI systems handle problems involving nuanced analysis and decision-making.
Applicants must hold or be completing a PhD in a relevant discipline or demonstrate equivalent research experience. Practical experience using Claude and Claude Code in their regular work is also required. Experience in developing evaluation frameworks, benchmarks or datasets is preferred.
The STEM Fellowship brings researchers from scientific and engineering disciplines into Anthropic to investigate challenging problems using Claude. Fellows will identify open research questions, create datasets and simulations, and design rigorous evaluations to determine where AI models fall short of expert-level reasoning.
Participants will also apply Claude to research problems and integrate relevant scientific tools and software into their investigations. Applicants must hold a PhD in a STEM field and possess several years of experience, such as postdoctoral research or faculty work. Previous machine learning experience is helpful but not compulsory.
The selection process includes a take-home assessment and discussions with researchers and potential mentors about applicants’ expertise, research interests and proposed projects. Anthropic expects to make offers by November 20, 2026.
The fellowships are full-time, with San Francisco as the default location. Remote participation may be considered in certain circumstances for candidates based in the United States. All fellows must attend a one-week in-person orientation in San Francisco.
Anthropic does not sponsor visas for these programmes. Applicants must already have, or independently obtain, authorisation to work full-time in the United States for the fellowship period. Interested candidates can review the requirements and apply through Anthropic’s official fellowship listings before the October 28 deadline.
