Research Scientist, Gemini (Remote — Worldwide)

Join the team behind Google’s most capable AI models and push the frontiers of multimodal large language modeling. As a Research Scientist for Gemini, you will contribute to the next generation of pre-training, fine-tuning, and reasoning capabilities that power global AI infrastructure. This role offers the unique opportunity to conduct world-class research at scale while operating in a fully remote environment across the globe.

- Design and implement novel architectures and training objectives for the Gemini family of multimodal models. - Conduct experiments to improve model reasoning, long-context understanding, and cross-modal integration. - Optimize large-scale training pipelines for performance, stability, and efficiency on TPU v5/v6 infrastructure. - Develop advanced fine-tuning and alignment techniques including RLHF, RLAIF, and supervised instruction tuning. - Analyze model failures and scaling laws to inform future architectural and data-driven iterations. - Collaborate with cross-functional teams in infrastructure, safety, and product to deploy Gemini at scale. - Contribute to the broader AI community through peer-reviewed research and technical documentation.

- PhD in Computer Science, Machine Learning, Mathematics, or a related quantitative field. - Proven track record of high-impact research publications at conferences such as NeurIPS, ICML, ICLR, or CVPR. - Extensive experience training and scaling Transformer-based architectures and large-scale language models. - Proficiency in Python and deep learning frameworks, specifically JAX and Flax. - Strong understanding of optimization techniques, reinforcement learning from human feedback (RLHF), and data curation strategies. - Experience working with distributed systems and large-scale TPU/GPU compute clusters. - Ability to collaborate effectively across time zones in a fast-paced, iterative research environment.

Top-tier compensation, equity, benefits.