Hattie Wu

Hattie is a performance optimization specialist with close to 10 years in machine learning engineering. She specializes in GPU platform performance tuning, covering full-stack optimization from workload analysis and framework customization to kernel and runtime acceleration, enabling end-to-end throughput and latency gains for AI systems.

Posts by Hattie Wu

July 08, 2026

SGLang-ATOM: Bring ROCm-Native Acceleration to SGLang Serving

Explore how SGLang-ATOM connects SGLang serving applications with ROCm-native ATOM execution to accelerate LLM inference on AMD Instinct GPUs.

https://rocm.blogs.amd.com/software-tools-optimization/atom-sglang-inference/README.html

June 29, 2026

Accelerating LLM Inference on AMD GPUs with Low-Latency GEMMs

Learn how FlyDSL low-latency GEMMs speed up LLM decode on AMD GPUs with Split-K, K-slice parallelism, and an LDS-based pipeline.

https://rocm.blogs.amd.com/software-tools-optimization/accelerating-llm-inference-on-amd-gpus-with-low-latency-gemms/README.html

June 15, 2026

ATOM: Unlocking Extreme AMD Instinct Inference with Software-Hardware Co-Optimization

A technical walkthrough of ATOM on AMD Instinct GPUs, covering architecture, feature scope, model coverage, and practical benchmark dashboard usage.

https://rocm.blogs.amd.com/software-tools-optimization/atom-inference-engine/README.html

May 07, 2026

vLLM-ATOM: Unlocking Native AMD Performance in the vLLM Ecosystem

Use ATOM as an out-of-tree vLLM plugin to keep vLLM compatibility while enabling AMD-optimized attention, model execution, and multi-model support including Kimi-K2.5.

https://rocm.blogs.amd.com/software-tools-optimization/vllm-atom/README.html