Fellowship•USA · Berkeley

Constellation Astra Fellowship

Funding

Fully Funded

Program dates

TBA

The Astra Fellowship is a fully funded, in-person program run by Constellation that pairs senior advisors and mentors with emerging talent to work on roughly 5-month projects aimed at reducing catastrophic risks from advanced AI. It runs from Constellation’s research center in Berkeley, California, a hub the organization describes as drawing 300+ weekly visitors. Fellows choose one of two streams: an Empirical Stream focused on machine-learning safety research (AI alignment, AI control, model evaluations, scalable oversight), and a Strategy & Governance Stream for those working on AI macrostrategy, policy, and field-building. The fellowship is designed to help participants transition into full-time AI-safety roles or launch their own safety-focused organizations; prior AI-safety experience is not required.

Benefits & funding

  • Monthly stipend of $8,400 for the duration of the program.
  • For Empirical Stream fellows, a research/compute budget of roughly $15,000 per fellow per month.
  • A $2,000 travel budget (cited on the Strategy & Governance stream page) to attend relevant workshops and conferences.
  • Dedicated workspace at the Berkeley research center alongside the wider AI-safety community.
  • Weekly mentorship from senior experts, plus research-management and placement support.
  • Visa support and guidance for international applicants.
  • Incubation support for fellows starting new safety-focused initiatives.

Eligibility

  • Open to applicants of any nationality; international applicants are supported through the visa process.
  • No specific degree, age, or minimum-experience requirement is published — the program admits people from diverse academic and professional backgrounds.
  • Applicants should be motivated to reduce catastrophic risks from advanced AI and bring technical or domain-specific experience relevant to the focus areas.
  • Empirical Stream: aimed at those who can do machine-learning research in core technical safety topics.
  • Strategy & Governance Stream: for applicants with strong familiarity with catastrophic AI risk who are highly agentic and able to work on difficult, unscoped problems; macrostrategy or policy experience is a plus.
  • Prior AI-safety experience is not required.

Application process

  1. 1
    Review the two streams (Empirical and Strategy & Governance) and decide which fits your background.
  2. 2
    While applications are between cycles, submit the expression-of-interest form to be notified when the next cohort opens.
  3. 3
    When a cohort is open, complete the online application form; shortlisted applicants are then assessed and acceptances are issued ahead of onboarding.

Program timeline

  1. Applications open (January 2026 cohort)

    Aug 28, 2025

  2. Application deadline (January 2026 cohort)

    Sep 26, 2025

  3. Acceptances sent

    Nov 6, 2025

  4. Program starts (January 2026 cohort)

    Jan 5, 2026

  5. Core program ends; optional extension period follows

    Mar 31, 2026

  6. Next cohort expected, per the Strategy & Governance stream page; submit the expression-of-interest form for exact dates.

    Fall 2026 (Sep 2026 – Feb 2027)