Principal AI Virtualization Engineer (GPU/XPU) - Equity

Broadcom Inc.; Austin, Northern (TX, KY); Hybrid; USD 122,000 - 195,000; Full time

Benefits offered by this job

Discretionary annual bonus

New hire equity grant

Annual equity awards

401(k) with company match

Employee Stock Purchase Program (ESPP)

Paid holidays

Paid time off

Job summary

Broadcom Inc. seeks a Principal Software Engineer to advance the AI Virtualization Stack for ESXi, enabling hardware-agnostic acceleration of AI/ML workloads in virtual machines.

You will develop PyTorch and JAX backends with OpenXLA, optimize ML acceleration, and collaborate with internal and external teams to support multi-vendor GPUs and XPUs. The role emphasizes performance tuning (KV-caching, FlashAttention) and requires strong C++/Python skills, plus experience with ML frameworks and

Qualifications

• 8+ years in ML framework/runtime development.
• Experience with C++ and Python.
• Strong knowledge of PyTorch and JAX.
• Familiarity with ML compilers and OpenXLA.
• Experience with version control systems like Git.

Responsibilities

• Research, design, and develop the AI Virtualization Stack for ESXi server development.
• Implement and optimize PyTorch and JAX backends using OpenXLA for high-performance AI/ML workloads across GPUs and XPUs.
• Analyze and re-architect performance-critical ML acceleration code, focusing on KV-caching and FlashAttention for LLM inference.
• Troubleshoot and resolve bugs in AI/ML acceleration functionality.
• Deliver code that meets Broadcom coding guidelines and document features.
• Collaborate with virtual driver and device teams and external GPU/XPU vendors for end-to-end ML framework support.

Skills

ML framework dev

PyTorch

JAX

OpenXLA

C++

Python

GPU kernel

Git

Education

Bachelor's degree in Computer Science

Master's degree in Computer Science

Tools

OpenXLA tooling

CUDA

ROCm

Git

Job description

Broadcom Inc. seeks a Principal Software Engineer to advance the AI Virtualization Stack for ESXi, enabling hardware-agnostic acceleration of AI/ML workloads in virtual machines.

You will develop PyTorch and JAX backends with OpenXLA, optimize ML acceleration, and collaborate with internal and external teams to support multi-vendor GPUs and XPUs. The role emphasizes performance tuning (KV-caching, FlashAttention) and requires strong C++/Python skills, plus experience with ML frameworks and