Research Scientist, Planning, Reasoning, Inference & Structured models (PRISM)

Snapshot

Artificial Intelligence could be one of humanity’s most useful inventions. At Google DeepMind, we’re a team of scientists, engineers, machine learning experts and more, working together to advance the state of the art in artificial intelligence. We use our technologies for widespread public benefit and scientific discovery, and collaborate with others on critical challenges, ensuring safety and ethics are the highest priority. We are looking for a scientist to become part of the PRISM team and join us in solving some of the hardest problems in AI.

About Us

The PRISM (Planning, Reasoning, Inference & Structured models) team is an interdisciplinary team which pioneers fundamental advances at the edge of large-scale artificial intelligence systems, building toward a future of autonomous agents with advanced reasoning and problem-solving capabilities.

We bridge the critical gap between groundbreaking research and real-world application. Our team is structured to do both: pursue high-risk, high-reward research into the core challenges of AI while simultaneously working as an integrated part of the core Gemini team to ship those innovations. 

We are currently focusing on developing advanced reasoning capabilities for Large Language Models, through both novel model improvements and innovative test-time techniques. As an integral part of the core Gemini team, we ensure our research and engineering work translates directly into shipped capabilities, making more powerful and useful intelligent systems widely available.

Beyond immediate product impact, we also aim to tackle AI grand challenges, such as developing training algorithms that can learn continuously and at scale on very general forms of data, to develop systems that do not just solve problems but are able to identify what the relevant problems to solve are, that learn to explore complex spaces and learn how to interact with rich environments. 

Our team's impact is demonstrated through core contributions to major Google initiatives. Recent highlights include:

• Gemini & Gemma: Developing core reasoning capabilities and pre-training architectures for Gemini 2.5 and spearheading the development of Gemma 3 270M.
• AI Grand Challenges: Making critical contributions that led to the gold medal-winning performance in the IMO 2025 effort.
• Product Innovation: Contributed to the model training of 'Deep Think' mode for Gemini, enabling more advanced problem-solving.
• Alphabet-Wide Collaboration: Driving key projects across Alphabet, including applications in AI for Science and enhancing cybersecurity through Project Big Sleep.

The Role

As a research scientist, you’ll be responsible for the following:

Key responsibilities:

• Define unsolved problems in the development of AI systems, propose solutions to tackle these issues, ranging from model architecture, data generation process, training algorithm or paradigm, and inference-time techniques.
• Carry out a rigorous process of implementation of these ideas in large-scale codebases on massive-compute infrastructure.
• Conduct thorough scientific analysis to ascertain the value of the proposed methods.
• Collaborate with partners and carry the most promising methods into production.
• Drive innovation from concept to implementation: Stay at the forefront of the AI field, identifying and exploring novel research avenues, implementing cutting-edge techniques, and validating their effectiveness through comprehensive experimental analysis.
• Provide technical guidance and mentorship to other engineers and researchers. 
• Take ownership of critical issues and drive collaborative efforts to ensure project success.

About You

In order to set you up for success as a Research Scientist at Google DeepMind,  we look for the following skills and experience:

• Possesses a deep intellectual curiosity and a relentless drive to understand the intricacies of machine learning models and training systems.
• A true team player who prioritizes collective success. This person is flexible, collaborative, and willing to contribute wherever needed to achieve project goals, fostering a positive and supportive team environment.
• Embraces a hacker mindset, fearlessly diving into complex codebases and technical challenges. Capable of both rapid prototyping for experimentation and developing robust, production-ready code for long-term solutions.
• Programming experience in Python and machine learning libraries such as jax, tensorflow or pytorch.
• Experience implementing state of the art machine learning algorithms and understand how to analyze results using notebooks and libraries such as pandas.
• Great communication skills, able to explain ideas at the conceptual level.

We are also interested in:

• A PhD degree in computer science, math or physics.
• Experience with serving and training Large Language Models.
• Knowledge of hardware acceleration of machine learning algorithms. 
• Knowledge of distributed computing. 
• Interest in reinforcement learning.