Google deepmind

Software Engineer, Numerics

Mountain View, California, US Full Time

Snapshot:

Artificial Intelligence could be one of humanity’s most useful inventions. At 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.

About Us:

This is a high impact role that will impact the efficiency of training Gemini through low precision fully quantized training as well as serving efficiency through very low precision and sparse models. Additionally, this role will drive the future Google HW roadmap to support forward looking numerics. Drive the vision and execution for quantization and sparsity efforts for Google DeepMind (GDM), including the flagship Gemini models, impacting both training and serving efficiency across Google. This role offers the unique opportunity to address a historically underserved but increasingly critical area in the advancement of AI.

The Role:

This high-impact position is responsible for deciding the precision, numerics, and sparsity formats used by Gemini and ensuring these decisions are reflected in the roadmaps for mobile and datacenter hardware, including TPUs. You will have the opportunity to influence the very architecture of Google's custom silicon for AI.

  • Decide the precision, numerics, and sparsity formats used by Gemini production models
    • Training on current and future Google HW platforms
    • Serving on a diverse set of platforms including TPU and Mobile
  • Drive novel research advancements and bringing the most impactful ideas into production
  • Ensure that Gemini numerics decisions are reflected in mobile and datacenter hardware roadmaps

About You:

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

  • PhD in Computer Science or related field with at least 2+ years of relevant experience.
  • Strong software-engineering skills in addition to a research background
  • Deep understanding of the numerics/quantization/sparsity literature 
  • Practical experience driving low precision and sparse models through to production