Posts by Arthur Huang

GEAK V3: Agent-Driven, Repository-Level GPU Kernel Optimization across HIP, Triton, and FlyDSL on AMD GPUs

• 20 July 2026
Saptarshi Majumder , Umang Pandey , Subrahmanya Pavankumar Dubagunta , Aditya Kumar Singh , Yue Liu , Chao Xu , Ethan Yang , Dan Li , Jianghui Wang , HongTao Meng , Puyuan Yang , Fan Wang , Bin Ding , Zheng Yao , Arthur Huang , Fuwei Yang , Manem Chaitanya , Johanna Xuyue Yang , Kristoffer Peyron , Balazs Toth , Yashvardhan Agarwal , Baiqiang Xia , Mehdi Saeedi , Ziqiong Liu , Arsalan Farooq , Pratik Prabhanjan Brahma , Dong Li , Zicheng Liu , Emad Barsoum
English
Applications & models
AI/ML GenAI LLM

In the ever-evolving world of GPU computing, optimizing kernels for performance and efficiency is a critical challenge. Hand-tuning kernels demands deep technical expertise and manual iteration. In this blog, you will read about how GEAK v3, the latest iteration of the agent-driven framework, tackles this problem using enhanced features such as task planning, test-harness discovery, patch-based handling of multi-file kernels, dynamic memory system and expert knowledge database. Our results show improvements across three kernel languages (HIP, Triton, and FlyDSL) and both CDNA and RDNA GPUs.

Read more ...

---

GEAK Agent-Driven Optimization of the DeepSeekV4 MLA Kernel

• 13 July 2026
Dan Li , Arthur Huang , Yue Liu , Ziqiong Liu , Dong Li , Emad Barsoum
English
Software tools & optimizations
AI/ML HPC GenAI LLM Performance

Optimizing LLM inference kernels requires more than a single kernel rewrite. Developers need to migrate reference implementations, analyze profiling results, validate correctness, and iterate quickly across different workload shapes. In this blog, we use DeepSeekV4 MLA as a case study to show how GEAK automates this workflow, from PyTorch-to-Triton migration to kernel-level optimization and SGLang end-to-end(E2E) validation on AMD GPUs.

Read more ...

---