UC Berkeley Is Hiring AI Researchers to Work on One of the Most Important Problems in Science Right Now
Stuart Russell has spent years arguing that the AI field is building increasingly powerful systems without adequately solving the question of how to make them do what humans actually want.
His 2019 book made the case. His TED talks have reached millions. And the research center he founded at UC Berkeley — the Center for Human-Compatible Artificial Intelligence — exists to work on that problem with the seriousness it deserves.
CHAI is now accepting applications for its CHAI Research Fellowship: a fully funded, two-year postdoctoral appointment at Berkeley with the possibility of renewal for a third year, visa sponsorship for international researchers and their families, and the freedom to design an independent research agenda within one of the most consequential fields in science today.
If AI alignment, machine learning, decision-making systems, robotics, or human-compatible AI is where your research sits — this is the fellowship worth knowing about.
CHAI Research Fellowship Overview
| Fellowship | CHAI Research Fellowship |
| Host Institution | UC Berkeley |
| Center | Center for Human-Compatible AI (CHAI) |
| Duration | 2 years (renewable up to 3 years) |
| Start Date | Flexible |
| Funding | Fully funded |
| Visa Sponsorship | Yes — including for dependents |
| Eligible Applicants | International researchers welcome |
| Requirement | PhD completed or near completion |
The Research Freedom Here Is Real
Many fellowships describe research independence in their materials and deliver something closer to a research assistant position in practice.
CHAI is structured differently. Fellows are expected to design and lead their own research agendas, not execute someone else’s.
That means proposing your own questions, building your own collaborations, and publishing your own findings — with access to CHAI’s faculty, BAIR’s ecosystem, and Berkeley’s broader academic infrastructure as resources rather than constraints.
The center’s current priority research areas are genuinely broad:
AI Foundations & Theory
- Theoretical foundations of intelligent agent architecture
- Formal reasoning systems
- Control theory and intelligent systems
- Probabilistic programming
Human-Centered AI
- Human-compatible AI systems
- Human preference learning
- Human-in-the-loop planning
- Human-robot interaction
Learning & Decision-Making
- Reinforcement learning
- Decision-making over long time horizons
- Cumulative lifelong learning
- Multi-agent cooperation
Alignment & Governance
- AI alignment and control
- Game theory applications in AI
- Mechanism design for human-machine systems
Researchers may also pursue interdisciplinary work spanning mathematics, economics, statistics, robotics, and cognitive science. The breadth is intentional — CHAI’s approach to alignment draws on multiple disciplines precisely because the problem resists single-discipline solutions.
The Funding Package
Two-year appointment, fully funded.
Fellows receive a competitive salary based on UC Berkeley’s established salary scales — which, at a major research university in a high cost-of-living area, are calibrated to be livable rather than merely symbolic.
Beyond the base funding:
- Renewal for a third year is possible based on performance
- Funding covers the full appointment period — not a declining award that phases out in year two
- Access to Berkeley’s research infrastructure, computing resources, and conference funding
For international researchers, CHAI adds something most fellowships don’t: visa sponsorship, including support for visa-related expenses. More on that in the next section.
The Visa Support — Read This Carefully
This detail is buried in most descriptions of the fellowship. It shouldn’t be.
CHAI provides:
- Visa sponsorship for successful international fellows
- Support for visa-related expenses — the costs that come with the sponsorship process
- Contributions toward visa costs for dependents — including partners and children
That last point is genuinely unusual. Most US research fellowships sponsor the researcher. The question of what happens to a partner or family member who relocates to Berkeley alongside them is usually left to the individual to navigate and fund.
CHAI explicitly addresses it. For researchers with families, or for those considering whether relocation to the United States is practically feasible, this is a meaningful difference between this fellowship and otherwise comparable opportunities.
Who Should Apply
The eligibility requirements are broader than you might expect.
The core requirement is a PhD — completed or near completion — in a relevant technical discipline. But CHAI’s definition of relevant is wider than most AI fellowships:
- Computer Science
- Statistics
- Mathematics
- Theoretical Economics
- Machine Learning
- Artificial Intelligence
- Robotics
- Data Science
And here’s the part worth highlighting for researchers who might otherwise rule themselves out: prior work in AI safety or alignment is not required.
CHAI is looking for strong researchers with solid technical foundations and genuine research potential. If your background is in reinforcement learning, game theory, statistical methods, cognitive modeling, or mechanism design — and your work touches on questions relevant to how AI systems interact with humans — you are likely eligible to apply, even if “AI alignment” has never appeared in your publication list.
Knowledge of modern AI and statistical learning methods is listed as an advantage, not a requirement.
Strong research achievements demonstrated through publications are expected. Leadership potential is assessed as part of the evaluation — CHAI fellows mentor junior researchers and graduate students, and the center looks for people who can grow into that role.
What the Application Requires
The application is deliberately lean. CHAI is not asking for a full research proposal or an extensive portfolio of documents.
Required materials:
- A current CV or résumé
- A one-page statement of interest outlining your proposed research direction
- Names and email addresses of two academic referees
That one-page statement carries significant weight. In a single page, you need to communicate what you want to work on, why it matters, why CHAI is the right environment for that work, and why you are the right person to do it.
A Few Things Worth Knowing Before You Apply
- You don’t need to have worked on AI safety before.
- The one-page statement is everything.
- Contact with faculty before applying helps.
- The flexible start date cuts both ways.
Official Link
UC Berkeley CHAI Research Fellowship
Related reading: