TL;DR

Tpu Kernel Engineer (AI): Designing and optimizing kernels for TPU to improve performance of large-scale ML systems including research, training, and inference with an accent on low-level optimization and system performance. Focus on debugging kernel performance at assembly level, implementing custom communication algorithms, and adapting models for low-precision inference.

Location: Hybrid with at least 25% office presence in San Francisco, USA

Salary: $280,000 - $560,000 USD annually

Company

%hirify_global% is a public benefit corporation focused on creating reliable, interpretable, and steerable AI systems to ensure AI safety and societal benefit.

What you will do

Identify and address performance issues across ML systems including research, training, and inference
Design and optimize kernels for TPU
Provide feedback to researchers on model performance impacts
Implement low-latency, high-throughput sampling for large language models
Build quantitative models of system performance and debug kernel performance at assembly level
Design and implement custom collective communication algorithms

Requirements

Must have at least a Bachelor's degree or equivalent experience
Hybrid work format requiring at least 25% office presence in San Francisco
Visa sponsorship available but not guaranteed for all candidates
Significant experience optimizing ML systems for TPUs, GPUs, or other accelerators
Strong track record in low-level optimization and large-scale systems problem solving
Good communication skills and collaborative mindset

Nice to have

Experience with ML framework internals
Background in computer architecture
Experience with language modeling and transformers

Culture & Benefits

Competitive compensation including equity and benefits
Generous vacation and parental leave
Flexible working hours
Supportive and collaborative research environment
Public benefit corporation with focus on ethical AI