Edge AI ML Kernel Performance Engineer

Join Amazon Devices as an ML Kernel Performance Engineer focused on Edge AI. Develop high-performance CUDA and Triton kernels to enhance model compression for efficient training and inference.
In this role, you will operate at the hardware-software boundary, collaborating with cross-functional teams to design and implement kernels for quantization-aware training and low-bit inference. Your work will involve analyzing kernel performance, identifying bottlenecks, and optimizing performance for cloud and edge deployments using modern GPU accelerators.

Key Responsibilities:

• Design CUDA and Triton kernels for efficient model training

• Conduct performance analysis to resolve training bottlenecks

• Implement kernel optimizations for compression tasks

• Create a kernel development harness for profiling

• Maintain a comprehensive training kernels library
Requirements:

• 3+ years in software development

• 2+ years in design or architecture of systems

• Experience with CUDA and ML kernels

• Proficient in Python, Java, C++

• Understanding of GPU memory hierarchies
Drive edge AI advancements with optimized kernel development at Amazon Devices.
#J-18808-Ljbffr