Google DeepMind is looking to solve the fundamental technical challenges required for the safe development of advanced AI systems, ensuring they are robust, predictable, and aligned with human values.
Requirements
- Experience in designing and implementing machine learning algorithms in Python.
- Hands-on experience with at least one major deep learning framework such as JAX, PyTorch, or TensorFlow.
- A strong mathematical background and the ability to engage with theoretical research in machine learning.
- Working experience with modern AI architectures (e.g., Transformers, Diffusion) and a comprehensive knowledge of machine learning fundamentals.
- Expertise in one or more of these areas: AI safety, model interpretability (explainable AI), or AI alignment.
- Experience with large-scale training and evaluation of deep learning models
Responsibilities
- Conceive, design, and lead research projects to develop novel algorithms for the safety, alignment, and interpretability of advanced AI agents.
- Drive the full research lifecycle, from theoretical ideation and mathematical formulation to implementation and large-scale empirical validation.
- Design and conduct rigorous evaluations and red-teaming exercises to discover and mitigate potential risks in next-generation models.
- Write software to implement research ideas and iterate quickly.
- Collaborate with world-class research and engineering teams to embed research directly into the AI development lifecycle.
- Stay at the forefront of the field by engaging with the latest theoretical research in machine learning and AI safety.
Other
- A PhD in Computer Science, Machine Learning, or a related technical field, or equivalent practical experience.
- A strong track record of research, including publications in top-tier peer-reviewed conferences or journals (e.g., NeurIPS, ICML, ICLR).
- Experience leading research projects and engaging in team collaborations to meet ambitious research goals.
- A strong interest in ensuring AI development benefits humanity.
- Ability to pass a background check
