Astra Fellowship 2027: Fully Funded 5-Month AI Safety Research Fellowship at Constellation in Berkeley
Overview
Astra is the flagship fellowship of Constellation, created to grow the pool of people working on AI safety. It is a fully funded, five-month, in-person programme held at Constellation's research center in Berkeley, California. Fellows work on frontier AI safety projects with expert mentors, dedicated research management and career support, in an environment centred on reducing risks from advanced AI.
Constellation is currently collecting expressions of interest so that candidates can be notified when the next cohort opens. According to Constellation, more than 80% of the first cohort now work full time in AI safety, at organisations including Redwood Research, METR, Anthropic, OpenAI, Google DeepMind, the Center for AI Standards and Innovation and the UK AI Security Institute. For the next round it aims to help more people move into high-impact roles and to establish new initiatives on urgent and neglected problems.
What you get
Fellows receive research opportunities, mentorship from experienced people in the field, research management, career support and connections with people and organisations working on AI safety. They are encouraged to pursue new ideas rather than follow a fixed assignment.
Who can apply
Constellation wants talented people who are motivated to reduce catastrophic risks from advanced AI. Applicants should bring technical or domain-specific experience relevant to AI safety, in the two focus areas of empirical research or strategy and governance. The programme is especially relevant to people who want to move into a full-time AI safety role or start an AI safety organisation.
Requirements
Previous experience specifically in AI safety is not required. Constellation notes that many impactful fellows have come from adjacent fields.
How to apply
Interested candidates can register an expression of interest to be notified when the next cohort launches. No deadline is given in the source.