Research Scientist, Multimodal Alignment, Safety, and Fairness
Remote • Kirkland, Washington, US; Mountain View, California, US; New York City, New York, US
Posted 3mo ago
Job Location
Kirkland, Washington, US; Mountain View, California, US; New York City, New York, US
Tech Stack
Remote Work Policy
Fully remote
Categories
AI Research Engineer
About the job
Google DeepMind's Frontier AI unit is seeking experienced Research Scientists to join a multimodal safety research effort. This role focuses on interdisciplinary sociotechnical modeling and requires a passion for understanding AI-society interactions, a strong awareness of AI alignment and safety, and a drive to develop novel ideas, methods, interfaces, and tools. You will contribute to advancing the state of the art in AI research and Google DeepMind's mission towards Artificial General Intelligence (AGI), with a focus on leading new breakthrough research directions in areas like AI behavior exploration, assessment, and steering, particularly for subjective and creative tasks. The work involves tackling fundamental research questions to improve alignment objectives, assess adherence to desired behaviors, and enable AI agents to monitor real-world social context and evolve system behaviors over long time-horizons. You will develop new paradigms for human+AI rating that are adaptive and context-aware, driving breakthroughs within Google DeepMind, Google products, and the broader AI alignment community.
Responsibilities
- Generate new research ideas and stay current with state-of-the-art advancements.
- Design, implement, and evaluate novel methods, interfaces, and tools for sociotechnical AI systems.
- Communicate research findings clearly and efficiently through presentations and written reports.
- Engage in inter- and intra-team collaborations to achieve research goals.
- Lead technical projects from ideation through implementation and evaluation.
- Develop new paradigms for human+AI rating systems.
- Research methods for AI agents to monitor social context and evolve system behaviors.
Requirements
- PhD degree in Computer Science, Machine Learning, or a related technical field.
- Strong publication record in top machine learning conferences.
- Hands-on experience developing multimodal AI models and systems.
- Strong programming skills in Python.
- Experience with at least one major deep learning framework (e.g., JAX/Flax/Gemax).
- Experience conducting independent research, including experimental design, implementation, and analysis.
- Expertise in working with and tuning large-scale vision language models (advantageous).
- Experience prototyping with VLMs using modern prompting strategies (advantageous).
- Experience fine-tuning and post-training LLMs using RL (advantageous).
- Experience developing agentic AI solutions to complex problems (advantageous).
- Awareness of the AI alignment landscape.