Join Google Cloud as a Senior Software Engineer focused on ML Compiler, Frameworks, and Performance. You will contribute to the evolution of machine learning infrastructure, driving innovation in areas such as distributed computing, large-scale system design, and artificial intelligence. This role offers the opportunity to work on critical projects within Google Cloud's Core ML organization, impacting Google's products and the broader ML community.

Google Cloud empowers organizations globally to digitally transform, leveraging cutting-edge technology and sustainable tools. We are seeking engineers who bring novel ideas and a passion for pushing technological boundaries across the full stack.

Drive continuous improvements across Google's ML software and hardware stack for internal teams and Google Cloud Platform customers. Develop a deep understanding of ML training and serving components, including frameworks like JAX and PyTorch, and the Accelerated Linear Algebra (XLA) runtime.

Identify and resolve performance bottlenecks in ML workloads and custom kernels. Collaborate with teams owning different parts of the ML stack, including open-source ML inference frameworks like TorchTPU, vLLM, and SGLang, to provide performance insights. Support new ML paradigms and architectures, such as horizontal scaling for upcoming TPU chips, through contributions to the stack and performance analysis tools.

Requires a Bachelor's degree or equivalent practical experience, along with a minimum of 5 years of software development experience in C++ or Python. Prior experience in machine learning (ML) infrastructure development or ML performance engineering is essential.

Preferred qualifications include experience with ML compilers and their internal workings, specifically writing compiler optimization passes. Familiarity with accelerator hardware architectures (TPUs/GPUs) and ML inference frameworks such as vLLM, SG Lang, or Pathways is also highly valued. Experience with ML frameworks like TensorFlow, JAX, PyTorch, or Keras is a plus.