Design and optimize high-performance kernels and custom operators for attention, MoE, GEMM, quantization and collective communication across NVIDIA, AMD and AWS Trainium. Improve LLM inference runtimes, develop distributed training/inference solutions at scale, use compilers and SDKs (Neuron, Torch Dynamo, PyTorch/XLA), contribute to open-source, and publish technical findings.

The GPU Cloud Platform Engineer designs and operates multi-cluster GPU infrastructures for AI workloads, ensuring performance and efficiency across cloud environments.