Design and build scalable software capabilities to manage the availability, scheduling, and reliability of the Cloud TPU Hypercomputer stack (VMs, Networking, Storage, GKE etc.).
Architect infrastructure solutions to ensure industry-leading availability guarantees for large-scale training and inference workloads.
Develop telemetry and tooling to establish service level objectives (SLO) and service level agreements (SLA), and to enable rapid debugging of complex infrastructure issues across the fleet.
Collaborate with platform, hardware, networking, and SRE teams to scale and manage accelerator capacity, including new TPU generations, ensure a seamless experience for customers.
Design and implement reliable ML infrastructure that enables training and serving cutting edge models at massive scale, troubleshoot complex distributed system issues across the stack (hardware, kernel, network), build the automation, tooling, and telemetry needed to turn operational findings into permanent software fixes and improved SLOs.
Minimum qualifications:
Bachelor’s degree or equivalent practical experience.
2 years of experience in backend Infrastructure development.
Experience in general purpose coding languages like C++, Go, or Python development.
Experience with algorithms, data structures, software development, and distributed computing.
Preferred qualifications:
Experience designing reliable, fault-tolerant and high performance distributed systems.
Experience with building cloud based services ideally with GCP.
Experience with large-scale distributed systems or Machine Learning (ML) systems (training and serving for computer vision, speech recognition, natural language processing, machine translation models).
Experience with reliability, large-scale distributed systems, Go, Google Cloud Platform, tensor processing unit (TPU), and service level objectives.
