---
format: "aidr-story-markdown/v1"
id: "aa1cf074c8668349e19a5828a0accbaab80941cbc4cf3fb906cfe9efadad303a"
canonical_url: "https://aidr.today/aa1cf074?lang=en"
title: "Mixture-of-Experts kernel engineering on Amazon SageMaker AI"
lang: "en"
requested_lang: "en"
available_langs: ["en","vi"]
translation_fallback: null
fallback_fields: []
published_at: "2026-10-13T15:17:09.000Z"
category: "Infra"
topics: ["amazon","inference"]
source_urls: ["https://aws.amazon.com/blogs/machine-learning/mixture-of-experts-kernel-engineering-on-amazon-sagemaker-ai/"]
summary: "AWS engineers detailed kernel optimizations for fused Mixture-of-Experts models on Amazon SageMaker AI, including Swap-AB for WGMMA utilization, Triton backend routing for FP8 quantization, and the elimination of Triton JIT recompilation. These combined optimizations deliver up to 10% higher throughput and up to 9% faster Time to First Token on H200 GPUs."
---

# Mixture\-of\-Experts kernel engineering on Amazon SageMaker AI

> [Open the canonical story](<https://aidr.today/aa1cf074?lang=en>)

**Published:** 2026-10-13T15:17:09.000Z
**Category:** Infra
**Topics:** amazon, inference

## Summary

AWS engineers detailed kernel optimizations for fused Mixture\-of\-Experts models on Amazon SageMaker AI, including Swap\-AB for WGMMA utilization, Triton backend routing for FP8 quantization, and the elimination of Triton JIT recompilation\. These combined optimizations deliver up to 10% higher throughput and up to 9% faster Time to First Token on H200 GPUs\.

## Sources

- [Story source](<https://aws.amazon.com/blogs/machine-learning/mixture-of-experts-kernel-engineering-on-amazon-sagemaker-ai/>)