Staff ML Systems Architect - Co-Design & TPU Performance

Google; Mountain View (CA); On-site; USD 207,000 - 300,000; Full time

Job summary

Google seeks a software engineer for the MARC team to advance TPU performance and open-source model mapping. You will shape pre-pilot strategies, implement high-performance modules, and coordinate with hardware, compiler, and cloud teams to drive 12+ month roadmaps.

You will define strategies for enabling open-source ML models to achieve TPU performance with internal models at launch, and lead cross-functional efforts in AI and infrastructure at scale.

Qualifications

• Bachelor's degree or equivalent practical experience.
• 8 years of experience programming in C++ or Python.
• 5 years of experience testing, and launching software products.
• 5 years of experience with performance, large-scale systems data analysis, visualization tools, or debugging.
• 3 years of experience with software design and architecture.

Responsibilities

• Deliver certified performance benchmarks six months pre-pilot as primary targets for pricing, capacity planning, and kernel optimization.
• Adapt emerging open-source models into TPU-native paired variants, including sparsity optimization, to demonstrate differentiated TPU performance.
• Ingest and optimize multi-turn agentic workflows, reasoning loops, and prompt and decode topologies prior to silicon tape-out.
• Author robust reference implementations using PyTorch, JAX, or Pallas that meet pre-pilot goals under compiler constraints.
• Partner with director and executive technical leadership across TPU Hardware, XLA Compiler, and Cloud AI Systems to steer multi-year roadmaps.

Skills

C++

Python

Software testing

Software architecture

Performance analysis

Education

Bachelor's degree

Tools

PyTorch

JAX

Pallas

Triton

CUDA

MLIR

XLA

HLO

PyTorch/XLA

Job description

Google seeks a software engineer for the MARC team to advance TPU performance and open-source model mapping. You will shape pre-pilot strategies, implement high-performance modules, and coordinate with hardware, compiler, and cloud teams to drive 12+ month roadmaps.

You will define strategies for enabling open-source ML models to achieve TPU performance with internal models at launch, and lead cross-functional efforts in AI and infrastructure at scale.