Senior ML Compiler Engineer - TPU Backends & Optimizations

Socket.dev; Sunnyvale (CA); On-site; USD 174,000 - 252,000; Full time

Job summary

Google Cloud seeks engineers to advance compiler optimizations and TPU backend features for large-scale workloads. You will help lower operations and develop optimizations across the compiler stack, leveraging TPU architectures for improved horizontal scaling and performance.

You will contribute to tooling that aids debugging and prevents scaling issues, enhancing the engineering experience across teams serving billions of users.

Qualifications

• Bachelor's degree in Computer Science or related field (or equivalent practical experience).
• 5 years of software development experience in Python and C++.
• 3 years in ML infrastructure or hardware accelerators (TPUs/GPUs).
• 2 years in a low-level systems programming language (C++).
• Experience in performance and compilers.

Responsibilities

• Deliver compiler parallelization features and optimization techniques for TPU backend for large-scale workloads.
• Contribute to collective operation lowering/implementation on TPU platform.
• Develop compiler optimization techniques at lower level and throughout the compiler stack.
• Analyze TPU architecture features and leverage them for optimal horizontal scaling.
• Build compiler tooling for debugging, scaling issues, and better engineering experience.

Skills

Python

C++

ML infrastructure

low-level programming

compiler optimization

Education

Bachelor's degree in Computer Science or related technical field

Tools

TensorFlow

JAX

PyTorch

TPUs

GPUs

Job description

Google Cloud seeks engineers to advance compiler optimizations and TPU backend features for large-scale workloads. You will help lower operations and develop optimizations across the compiler stack, leveraging TPU architectures for improved horizontal scaling and performance.

You will contribute to tooling that aids debugging and prevents scaling issues, enhancing the engineering experience across teams serving billions of users.