Drive multi-quarter resource forecasting and allocation for machine learning hardware (TPUs and GPUs), host and storage systems, and cloud infrastructure to support scaling initiatives. Partner closely with research and engineering leads to model operational trade-offs and distribute compute capacity efficiently across pre-training, post-training, evaluations, and serving priorities. Coordinate end-to-end compute readiness and seamless cluster transitions for flagship training runs while actively managing dedicated reservations and dynamic quota pools.
Monitor and optimize overall fleet occupancy, hardware resource availability, scaling efficiency, and workload performance to rapidly resolve scheduling bottlenecks. Develop robust self-serve tooling, including SQL and Python data pipelines, live operational dashboards, actionable playbooks, and data-backed recommendations for leadership.
Job description provided by Google.
View the original posting
