Research Engineer at Aisafety

We are the Center for AI Safety, a non-profit research organization in San Francisco working to reduce high-consequence risks from advanced AI through technical research, community building, and collaboration with leading academics and policymakers.

Responsibilities

• Design and run experiments on large language models, including fine-tuning transformers and evaluating model behavior across varied data domains.
• Develop new datasets and benchmarks to measure model robustness, transparency, and safety properties.
• Evaluate models in sequential decision-making environments and other complex settings.
• Write and submit research papers to top-tier machine learning conferences.
• Organize and lead workshops, competitions, and community events that help grow the AI safety field.
• Collaborate with internal researchers and external academics at institutions such as Stanford, UC Berkeley, CMU, and MIT.

Must-haves

• Degree in computer science or a closely related machine learning field.
• Co-author on an NLP paper accepted to a top conference.
• Strong alignment with the mission of reducing societal-scale risks from AI.
• Ability to read and synthesize machine learning papers, placing new results in the context of existing literature.
• Proficiency with PyTorch and Hugging Face libraries.
• Experience launching and managing distributed machine learning jobs across hundreds or thousands of GPUs.
• Clear, timely communication with both technical and non-technical teammates.
• English fluency.

Nice-to-haves

• First-author publication at a top NLP or machine learning conference.
• Well-developed perspective on the most tractable, high-impact technical problems in AI safety.
• Experience organizing or leading research competitions, workshops, or similar community initiatives.

Benefits

• Health, dental, and vision insurance with 100% coverage for employees and coverage for dependents.
• 401(k) plan with 4% employer match.
• Competitive paid time off.
• Commuter benefits.
• Lunch and dinner provided at the office.
• Personalized ergonomic workstation setup.
• Direct access to leading researchers working on technical and conceptual AI safety problems.