Senior Technical Leader- Physical AI Software Development (ROCm)

Role Summary

As an Lead Engineer in AMD's Physical AI and ROCm Software teams, you will design, develop, and optimize systemlevel and runtime software that powers embodied and physicalworld AI workloads (robotics, perception, vision-language-action models, multimodal inference) on AMD accelerators including Ryzen AI NPUs, GPUs (Instinct/Radeon), and ROCm platforms.
You'll work on the intersection of AI system software, hardware acceleration, robotics/physical AI, and highperformance runtime frameworks.

This role combines responsibilities typically seen in ROCm HIP Runtime development, AI system software optimization, embedded AI enablement, and performance engineering.

Key Responsibilities

1. ROCm / HIP Runtime & System Software Development

Based on AMD's Lead HIP Runtime and ROCm AI system software roles:

• Design and develop new ROCm/HIP runtime features supporting AI workloads.
• Optimize the AI software stack across PyTorch, vLLM, GGML/llama.cpp, ONNX Runtime, and other opensource frameworks.
• Participate in bringup of platforms and enable AI acceleration on GPUs and NPUs.
• Implement highperformance kernels/operators and contribute to compiler/runtime optimization.
• Work closely with AMD architecture teams to guide future product enhancements.

2. Physical AI / Robotics Software Enablement

Relevant to Physical AI initiatives from internal decks:

• Develop system software enabling perception, navigation, robotics stacks, VLA (VisionLanguageAction) models, and embodied AI workloads on AMD ROCmbased systems.
• Optimize performance for robotics pipelines, sensor fusion, simulation, and control loops running on GPUs/NPUs.
• Support Physical AI model enablement with internal teams and external partners.
• Work on endtoend physicalAI inference pipelines including model deployment, quantization, and realtime optimization.

3. Performance Engineering & Debugging

• Conduct profiling, tuning, and performance bottleneck analysis for physicalAI and robotics workloads on AMD hardware.
• Develop/debug tools for runtime and modellevel optimization (profilers, visualizers, instrumentation).
• Provide deepdive debugging support across the AI software stack and runtime layers.

4. CrossFunctional Collaboration

• Collaborate with hardware, compiler, ML framework, robotics, and embedded Linux teams.
• Engage with internal teams and external partners
• Contribute to opensource ROCm components and partner with ecosystem developers.

Required Skills & Qualifications

Core Technical Skills

• Strong expertise in C/C++, with experience in performancecritical system software.
• Experience with GPU/NPU compute programming: HIP, CUDA, OpenCL is a strong plus.
• Deep understanding of AI/ML inference, quantization, kernel optimization, and runtime frameworks.
• Strong experience with Linux systems, build tools, debuggers, profilers, and source control (Git).
• Familiarity with robotics, perception stacks, simulation frameworks, or embedded AI systems is desirable.

Desirable Experience

• Experience with ROCm ecosystem, HIP runtime, MLIR, compilers, or driverlevel development.
• Knowledge of physical AI workloads such as multimodal models, depth/vision inference, navigation, etc.
• Exposure to embedded Linux, containers, and edge inference platforms.

Soft Skills

• Ability to drive sophisticated technical issues to closure independently.
• Strong communication and crossteam collaboration skills.
• Leadership mindset to influence roadmap and architecture decisions.

Education

• Bachelor's or Master's degree in CS, CE, EE, or equivalent.

Why This Role Is Unique

• Direct impact on AMD's Physical AI strategy, enabling embodied AI and nextgeneration robotics platforms.
• Work across the full AMD AI stack: NPUs, GPUs, compilers, runtimes, models, and robotics frameworks.
• Contribute to cuttingedge ROCm advancements and future AI hardware/software codesign.

#LI-PK1