EdgeTPU ML Compiler Engineer — Build High-Performance ML

Google; Mountain View (CA); On-site; USD 174,000 - 252,000; Full time

Benefits offered by this job

Health insurance

Dental insurance

Vision insurance

Disability insurance

401(k)

Job summary

Google is seeking a Senior Software Engineer to join the EdgeTPU Compiler team, focusing on next-generation compiler optimizations for ML models on custom hardware. You will translate framework-level code into efficient EdgeTPU instructions and own end-to-end features.

You will triage performance and correctness issues, collaborate with ML model owners to optimize architectures for EdgeTPU, and drive production-grade compiler improvements across projects.

Qualifications

• Bachelor’s degree or equivalent practical experience.
• 5 years of software development experience in one or more languages (e.g., C++).
• 3 years testing, maintaining, or launching software products, plus 1 year in design/architecture.
• 3 years with compilers (optimization and parallelization).

Responsibilities

• Design and build ML compiler optimizations for EdgeTPU hardware, and extend leading authoring frameworks (such as JAX and PyTorch) to enable seamless, high-performance compilation.
• Triage, root-cause, and resolve complex correctness and performance issues encountered when deploying state-of-the-art ML models on EdgeTPU.
• Propose, design, and implement robust compiler features and fixes to systematically address performance bottlenecks and hardware limitations.
• Partner closely with ML model owners to influence model architectures, ensuring they are designed for optimal, efficient execution on EdgeTPU systems.
• Own project execution from end to end, collaborating with cross-functional partners and stakeholders, and managing priorities/deadlines/deliverables for key compiler feature areas.

Skills

C++

Software development

Testing & QA

Compiler knowledge

Education

Bachelor's degree

Master's degree

PhD

Tools

JAX

PyTorch

Job description

Google is seeking a Senior Software Engineer to join the EdgeTPU Compiler team, focusing on next-generation compiler optimizations for ML models on custom hardware. You will translate framework-level code into efficient EdgeTPU instructions and own end-to-end features.

You will triage performance and correctness issues, collaborate with ML model owners to optimize architectures for EdgeTPU, and drive production-grade compiler improvements across projects.