AI-Driven ML Performance Engineer Intern - Hybrid/Remote

Waymo LLC; Mountain View (CA); Híbrido; USD 55.000 - 83.000; A tiempo parcial

Ventajas ofrecidas por este puesto de trabajo

Hybrid work model with remote work

Educational reimbursement

On-site meals and snacks

Competitive compensation

Descripción de la vacante

Waymo LLC is seeking a highly capable intern to design and optimize kernels for a transformer inference workload on Waymo’s ML accelerators. You will work with AI agents to push the performance of a Domain-Specific Language for a custom accelerator, ensuring correctness and numerical accuracy throughout development.

You will analyze bottlenecks, validate results, and build tooling to automate kernel generation, enabling iterative improvements with minimal human intervention.

Formación

• Working knowledge of kernels for transformer inference and performance analysis.
• Familiarity with low-level ML accelerator DSLs (CUDA, Triton, Gluon).
• Fluency in Python and/or C++.
• Currently enrolled in an MS or PhD program in Computer Science, Computer Engineering, Computational Science, or related quantitative field.

Responsabilidades

• Collaborate with AI agents to develop high-performance low-level kernels for a custom ML accelerator DSL.
• Ensure functional correctness and numerical accuracy of kernels.
• Conduct performance analyses to identify bottlenecks and optimization opportunities.
• Develop a harness to iteratively improve kernel generation and reduce manual tuning.

Conocimientos

Python

C++

Kernel optimization

Performance analysis

AI agent harnesses

Educación

MS or PhD in Computer Science/Computer Engineering/Computational Science

Herramientas

CUDA

Triton

Gluon

Descripción del empleo

Waymo LLC is seeking a highly capable intern to design and optimize kernels for a transformer inference workload on Waymo’s ML accelerators. You will work with AI agents to push the performance of a Domain-Specific Language for a custom accelerator, ensuring correctness and numerical accuracy throughout development.

You will analyze bottlenecks, validate results, and build tooling to automate kernel generation, enabling iterative improvements with minimal human intervention.