Senior Compiler Engineer - TPU & ML Systems

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

Job summary

Google is seeking software engineers to build the Accelerated Linear Algebra (XLA) compiler for TPU workloads. You will work on a state-of-the-art compiler enabling TPUs to accelerate ML and scientific workloads, with opportunities to collaborate across HW/SW co-design.

You will design, develop, test, and deploy scalable compiler solutions, optimize for new models and generations of TPU hardware, and contribute to Google Cloud and internal users’ computing capabilities.

Qualifications

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

Responsibilities

• Contribute to the compiler for a novel processor designed to accelerate machine learning workloads.
• Target and compile high-performance implementations of operations at distributed scale.
• Design and implement new compiler passes that extract more performance out of current and next-generation TPUs, directly impacting fleet efficiency.
• Collaborate closely with hardware designers to co-design future processors.
• Research high-level representations to effectively program large-scale, distributed, and heterogeneous systems.

Skills

C++/Python programming

Software testing & release

Performance & large-scale analysis

Software design & architecture

ML compiler knowledge

Education

Bachelor's degree or equivalent practical experience

Job description

Google is seeking software engineers to build the Accelerated Linear Algebra (XLA) compiler for TPU workloads. You will work on a state-of-the-art compiler enabling TPUs to accelerate ML and scientific workloads, with opportunities to collaborate across HW/SW co-design.

You will design, develop, test, and deploy scalable compiler solutions, optimize for new models and generations of TPU hardware, and contribute to Google Cloud and internal users’ computing capabilities.