<rss version="2.0" xmlns:atom="http://www.w3.org/2005/Atom"><channel><title>makgol's note – ML 플랫폼·인프라</title><link>https://docs.makgol.com/external/uber/ml-platform/</link><description>Recent content in ML 플랫폼·인프라 on makgol's note</description><generator>Hugo -- gohugo.io</generator><language>ko-KR</language><copyright>© 2026 Mont</copyright><lastBuildDate>Tue, 10 Nov 2026 00:00:00 +0900</lastBuildDate><atom:link href="https://docs.makgol.com/external/uber/ml-platform/index.xml" rel="self" type="application/rss+xml"/><item><title>GPU 메모리부터 줄인 Uber Michelangelo의 LLM 학습</title><link>https://docs.makgol.com/external/uber/ml-platform/open-source-and-in-house-how-uber-optimizes-llm-training/</link><pubDate>Tue, 10 Nov 2026 00:00:00 +0900</pubDate><guid>https://docs.makgol.com/external/uber/ml-platform/open-source-and-in-house-how-uber-optimizes-llm-training/</guid><description>
&lt;h1&gt;GPU 메모리부터 줄인 Uber Michelangelo의 LLM 학습 (2026-04)&lt;/h1&gt;&lt;div class="hx:overflow-x-auto hx:mt-6 hx:flex hx:rounded-lg hx:border hx:py-2 hx:ltr:pr-4 hx:rtl:pl-4 hx:contrast-more:border-current hx:contrast-more:dark:border-current hx:border-blue-200 hx:bg-blue-100 hx:text-blue-900 hx:dark:border-blue-200/30 hx:dark:bg-blue-900/30 hx:dark:text-blue-200"&gt;
&lt;div class="hx:ltr:pl-3 hx:ltr:pr-2 hx:rtl:pr-3 hx:rtl:pl-2"&gt;&lt;svg height=1.2em class="hx:inline-block hx:align-middle" xmlns="http://www.w3.org/2000/svg" fill="none" viewBox="0 0 24 24" stroke-width="2" stroke="currentColor" aria-hidden="true"&gt;&lt;path stroke-linecap="round" stroke-linejoin="round" d="M13 16h-1v-4h-1m1-4h.01M21 12a9 9 0 11-18 0 9 9 0 0118 0z"/&gt;&lt;/svg&gt;&lt;/div&gt;
&lt;div class="hx:w-full hx:min-w-0 hx:leading-7"&gt;
&lt;div class="hx:mt-6 hx:leading-7 hx:first:mt-0"&gt;&lt;p&gt;&lt;strong&gt;원문 정보&lt;/strong&gt;&lt;/p&gt;</description></item><item><title>GPU의 긴 기다림을 줄인 Petastorm 데이터 로더</title><link>https://docs.makgol.com/external/uber/ml-platform/accelerating-deep-learning/</link><pubDate>Tue, 10 Nov 2026 00:00:00 +0900</pubDate><guid>https://docs.makgol.com/external/uber/ml-platform/accelerating-deep-learning/</guid><description>
&lt;h1&gt;GPU의 긴 기다림을 줄인 Petastorm 데이터 로더 (2026-04)&lt;/h1&gt;&lt;div class="hx:overflow-x-auto hx:mt-6 hx:flex hx:rounded-lg hx:border hx:py-2 hx:ltr:pr-4 hx:rtl:pl-4 hx:contrast-more:border-current hx:contrast-more:dark:border-current hx:border-blue-200 hx:bg-blue-100 hx:text-blue-900 hx:dark:border-blue-200/30 hx:dark:bg-blue-900/30 hx:dark:text-blue-200"&gt;
&lt;div class="hx:ltr:pl-3 hx:ltr:pr-2 hx:rtl:pr-3 hx:rtl:pl-2"&gt;&lt;svg height=1.2em class="hx:inline-block hx:align-middle" xmlns="http://www.w3.org/2000/svg" fill="none" viewBox="0 0 24 24" stroke-width="2" stroke="currentColor" aria-hidden="true"&gt;&lt;path stroke-linecap="round" stroke-linejoin="round" d="M13 16h-1v-4h-1m1-4h.01M21 12a9 9 0 11-18 0 9 9 0 0118 0z"/&gt;&lt;/svg&gt;&lt;/div&gt;
&lt;div class="hx:w-full hx:min-w-0 hx:leading-7"&gt;
&lt;div class="hx:mt-6 hx:leading-7 hx:first:mt-0"&gt;&lt;p&gt;&lt;strong&gt;원문 정보&lt;/strong&gt;&lt;/p&gt;</description></item><item><title>수천 개 모델의 배포를 지키는 Michelangelo 안전장치</title><link>https://docs.makgol.com/external/uber/ml-platform/raising-the-bar-on-ml-model-deployment-safety/</link><pubDate>Tue, 10 Nov 2026 00:00:00 +0900</pubDate><guid>https://docs.makgol.com/external/uber/ml-platform/raising-the-bar-on-ml-model-deployment-safety/</guid><description>
&lt;h1&gt;수천 개 모델의 배포를 지키는 Michelangelo 안전장치 (2025-10)&lt;/h1&gt;&lt;div class="hx:overflow-x-auto hx:mt-6 hx:flex hx:rounded-lg hx:border hx:py-2 hx:ltr:pr-4 hx:rtl:pl-4 hx:contrast-more:border-current hx:contrast-more:dark:border-current hx:border-blue-200 hx:bg-blue-100 hx:text-blue-900 hx:dark:border-blue-200/30 hx:dark:bg-blue-900/30 hx:dark:text-blue-200"&gt;
&lt;div class="hx:ltr:pl-3 hx:ltr:pr-2 hx:rtl:pr-3 hx:rtl:pl-2"&gt;&lt;svg height=1.2em class="hx:inline-block hx:align-middle" xmlns="http://www.w3.org/2000/svg" fill="none" viewBox="0 0 24 24" stroke-width="2" stroke="currentColor" aria-hidden="true"&gt;&lt;path stroke-linecap="round" stroke-linejoin="round" d="M13 16h-1v-4h-1m1-4h.01M21 12a9 9 0 11-18 0 9 9 0 0118 0z"/&gt;&lt;/svg&gt;&lt;/div&gt;
&lt;div class="hx:w-full hx:min-w-0 hx:leading-7"&gt;
&lt;div class="hx:mt-6 hx:leading-7 hx:first:mt-0"&gt;&lt;p&gt;&lt;strong&gt;원문 정보&lt;/strong&gt;&lt;/p&gt;</description></item><item><title>클러스터 선택을 플랫폼에 맡긴 Ray on Kubernetes</title><link>https://docs.makgol.com/external/uber/ml-platform/ubers-journey-to-ray-on-kubernetes-ray-setup/</link><pubDate>Tue, 10 Nov 2026 00:00:00 +0900</pubDate><guid>https://docs.makgol.com/external/uber/ml-platform/ubers-journey-to-ray-on-kubernetes-ray-setup/</guid><description>
&lt;h1&gt;클러스터 선택을 플랫폼에 맡긴 Ray on Kubernetes (2025-04)&lt;/h1&gt;&lt;div class="hx:overflow-x-auto hx:mt-6 hx:flex hx:rounded-lg hx:border hx:py-2 hx:ltr:pr-4 hx:rtl:pl-4 hx:contrast-more:border-current hx:contrast-more:dark:border-current hx:border-blue-200 hx:bg-blue-100 hx:text-blue-900 hx:dark:border-blue-200/30 hx:dark:bg-blue-900/30 hx:dark:text-blue-200"&gt;
&lt;div class="hx:ltr:pl-3 hx:ltr:pr-2 hx:rtl:pr-3 hx:rtl:pl-2"&gt;&lt;svg height=1.2em class="hx:inline-block hx:align-middle" xmlns="http://www.w3.org/2000/svg" fill="none" viewBox="0 0 24 24" stroke-width="2" stroke="currentColor" aria-hidden="true"&gt;&lt;path stroke-linecap="round" stroke-linejoin="round" d="M13 16h-1v-4h-1m1-4h.01M21 12a9 9 0 11-18 0 9 9 0 0118 0z"/&gt;&lt;/svg&gt;&lt;/div&gt;
&lt;div class="hx:w-full hx:min-w-0 hx:leading-7"&gt;
&lt;div class="hx:mt-6 hx:leading-7 hx:first:mt-0"&gt;&lt;p&gt;&lt;strong&gt;원문 정보&lt;/strong&gt;&lt;/p&gt;</description></item><item><title>모델 개발의 길을 다시 설계한 Michelangelo의 8년</title><link>https://docs.makgol.com/external/uber/ml-platform/from-predictive-to-generative-ai/</link><pubDate>Tue, 10 Nov 2026 00:00:00 +0900</pubDate><guid>https://docs.makgol.com/external/uber/ml-platform/from-predictive-to-generative-ai/</guid><description>
&lt;h1&gt;모델 개발의 길을 다시 설계한 Michelangelo의 8년 (2024-05)&lt;/h1&gt;&lt;div class="hx:overflow-x-auto hx:mt-6 hx:flex hx:rounded-lg hx:border hx:py-2 hx:ltr:pr-4 hx:rtl:pl-4 hx:contrast-more:border-current hx:contrast-more:dark:border-current hx:border-blue-200 hx:bg-blue-100 hx:text-blue-900 hx:dark:border-blue-200/30 hx:dark:bg-blue-900/30 hx:dark:text-blue-200"&gt;
&lt;div class="hx:ltr:pl-3 hx:ltr:pr-2 hx:rtl:pr-3 hx:rtl:pl-2"&gt;&lt;svg height=1.2em class="hx:inline-block hx:align-middle" xmlns="http://www.w3.org/2000/svg" fill="none" viewBox="0 0 24 24" stroke-width="2" stroke="currentColor" aria-hidden="true"&gt;&lt;path stroke-linecap="round" stroke-linejoin="round" d="M13 16h-1v-4h-1m1-4h.01M21 12a9 9 0 11-18 0 9 9 0 0118 0z"/&gt;&lt;/svg&gt;&lt;/div&gt;
&lt;div class="hx:w-full hx:min-w-0 hx:leading-7"&gt;
&lt;div class="hx:mt-6 hx:leading-7 hx:first:mt-0"&gt;&lt;p&gt;&lt;strong&gt;원문 정보&lt;/strong&gt;&lt;/p&gt;</description></item><item><title>우버 AI/ML 인프라에서 찾은 GPU 효율의 조건</title><link>https://docs.makgol.com/external/uber/ml-platform/scaling-ai-ml-infrastructure-at-uber/</link><pubDate>Tue, 10 Nov 2026 00:00:00 +0900</pubDate><guid>https://docs.makgol.com/external/uber/ml-platform/scaling-ai-ml-infrastructure-at-uber/</guid><description>
&lt;h1&gt;우버 AI/ML 인프라에서 찾은 GPU 효율의 조건 (2024-03)&lt;/h1&gt;&lt;div class="hx:overflow-x-auto hx:mt-6 hx:flex hx:rounded-lg hx:border hx:py-2 hx:ltr:pr-4 hx:rtl:pl-4 hx:contrast-more:border-current hx:contrast-more:dark:border-current hx:border-blue-200 hx:bg-blue-100 hx:text-blue-900 hx:dark:border-blue-200/30 hx:dark:bg-blue-900/30 hx:dark:text-blue-200"&gt;
&lt;div class="hx:ltr:pl-3 hx:ltr:pr-2 hx:rtl:pr-3 hx:rtl:pl-2"&gt;&lt;svg height=1.2em class="hx:inline-block hx:align-middle" xmlns="http://www.w3.org/2000/svg" fill="none" viewBox="0 0 24 24" stroke-width="2" stroke="currentColor" aria-hidden="true"&gt;&lt;path stroke-linecap="round" stroke-linejoin="round" d="M13 16h-1v-4h-1m1-4h.01M21 12a9 9 0 11-18 0 9 9 0 0118 0z"/&gt;&lt;/svg&gt;&lt;/div&gt;
&lt;div class="hx:w-full hx:min-w-0 hx:leading-7"&gt;
&lt;div class="hx:mt-6 hx:leading-7 hx:first:mt-0"&gt;&lt;p&gt;&lt;strong&gt;원문 정보&lt;/strong&gt;&lt;/p&gt;</description></item><item><title>운영 모델의 품질을 지키는 Model Excellence Scores</title><link>https://docs.makgol.com/external/uber/ml-platform/enhancing-the-quality-of-machine-learning-systems-at-scale/</link><pubDate>Tue, 10 Nov 2026 00:00:00 +0900</pubDate><guid>https://docs.makgol.com/external/uber/ml-platform/enhancing-the-quality-of-machine-learning-systems-at-scale/</guid><description>
&lt;h1&gt;운영 모델의 품질을 지키는 Model Excellence Scores (2024-03)&lt;/h1&gt;&lt;div class="hx:overflow-x-auto hx:mt-6 hx:flex hx:rounded-lg hx:border hx:py-2 hx:ltr:pr-4 hx:rtl:pl-4 hx:contrast-more:border-current hx:contrast-more:dark:border-current hx:border-blue-200 hx:bg-blue-100 hx:text-blue-900 hx:dark:border-blue-200/30 hx:dark:bg-blue-900/30 hx:dark:text-blue-200"&gt;
&lt;div class="hx:ltr:pl-3 hx:ltr:pr-2 hx:rtl:pr-3 hx:rtl:pl-2"&gt;&lt;svg height=1.2em class="hx:inline-block hx:align-middle" xmlns="http://www.w3.org/2000/svg" fill="none" viewBox="0 0 24 24" stroke-width="2" stroke="currentColor" aria-hidden="true"&gt;&lt;path stroke-linecap="round" stroke-linejoin="round" d="M13 16h-1v-4h-1m1-4h.01M21 12a9 9 0 11-18 0 9 9 0 0118 0z"/&gt;&lt;/svg&gt;&lt;/div&gt;
&lt;div class="hx:w-full hx:min-w-0 hx:leading-7"&gt;
&lt;div class="hx:mt-6 hx:leading-7 hx:first:mt-0"&gt;&lt;p&gt;&lt;strong&gt;원문 정보&lt;/strong&gt;&lt;/p&gt;</description></item><item><title>Tier1 서빙 장애에서 출발한 Palette 메타 스토어 재설계</title><link>https://docs.makgol.com/external/uber/ml-platform/palette-meta-store-journey/</link><pubDate>Tue, 10 Nov 2026 00:00:00 +0900</pubDate><guid>https://docs.makgol.com/external/uber/ml-platform/palette-meta-store-journey/</guid><description>
&lt;h1&gt;Tier1 서빙 장애에서 출발한 Palette 메타 스토어 재설계 (2024-01)&lt;/h1&gt;&lt;div class="hx:overflow-x-auto hx:mt-6 hx:flex hx:rounded-lg hx:border hx:py-2 hx:ltr:pr-4 hx:rtl:pl-4 hx:contrast-more:border-current hx:contrast-more:dark:border-current hx:border-blue-200 hx:bg-blue-100 hx:text-blue-900 hx:dark:border-blue-200/30 hx:dark:bg-blue-900/30 hx:dark:text-blue-200"&gt;
&lt;div class="hx:ltr:pl-3 hx:ltr:pr-2 hx:rtl:pr-3 hx:rtl:pl-2"&gt;&lt;svg height=1.2em class="hx:inline-block hx:align-middle" xmlns="http://www.w3.org/2000/svg" fill="none" viewBox="0 0 24 24" stroke-width="2" stroke="currentColor" aria-hidden="true"&gt;&lt;path stroke-linecap="round" stroke-linejoin="round" d="M13 16h-1v-4h-1m1-4h.01M21 12a9 9 0 11-18 0 9 9 0 0118 0z"/&gt;&lt;/svg&gt;&lt;/div&gt;
&lt;div class="hx:w-full hx:min-w-0 hx:leading-7"&gt;
&lt;div class="hx:mt-6 hx:leading-7 hx:first:mt-0"&gt;&lt;p&gt;&lt;strong&gt;원문 정보&lt;/strong&gt;&lt;/p&gt;</description></item><item><title>조용한 데이터 누락을 찾아내는 Uber의 자동 감시 시스템 D3</title><link>https://docs.makgol.com/external/uber/ml-platform/d3-an-automated-system-to-detect-data-drifts/</link><pubDate>Tue, 10 Nov 2026 00:00:00 +0900</pubDate><guid>https://docs.makgol.com/external/uber/ml-platform/d3-an-automated-system-to-detect-data-drifts/</guid><description>
&lt;h1&gt;조용한 데이터 누락을 찾아내는 Uber의 자동 감시 시스템 D3 (2023-02)&lt;/h1&gt;&lt;div class="hx:overflow-x-auto hx:mt-6 hx:flex hx:rounded-lg hx:border hx:py-2 hx:ltr:pr-4 hx:rtl:pl-4 hx:contrast-more:border-current hx:contrast-more:dark:border-current hx:border-blue-200 hx:bg-blue-100 hx:text-blue-900 hx:dark:border-blue-200/30 hx:dark:bg-blue-900/30 hx:dark:text-blue-200"&gt;
&lt;div class="hx:ltr:pl-3 hx:ltr:pr-2 hx:rtl:pr-3 hx:rtl:pl-2"&gt;&lt;svg height=1.2em class="hx:inline-block hx:align-middle" xmlns="http://www.w3.org/2000/svg" fill="none" viewBox="0 0 24 24" stroke-width="2" stroke="currentColor" aria-hidden="true"&gt;&lt;path stroke-linecap="round" stroke-linejoin="round" d="M13 16h-1v-4h-1m1-4h.01M21 12a9 9 0 11-18 0 9 9 0 0118 0z"/&gt;&lt;/svg&gt;&lt;/div&gt;
&lt;div class="hx:w-full hx:min-w-0 hx:leading-7"&gt;
&lt;div class="hx:mt-6 hx:leading-7 hx:first:mt-0"&gt;&lt;p&gt;&lt;strong&gt;원문 정보&lt;/strong&gt;&lt;/p&gt;</description></item><item><title>모델 배포와 서비스 릴리스를 분리한 Uber의 ML 서빙</title><link>https://docs.makgol.com/external/uber/ml-platform/continuous-integration-deployment-ml/</link><pubDate>Tue, 10 Nov 2026 00:00:00 +0900</pubDate><guid>https://docs.makgol.com/external/uber/ml-platform/continuous-integration-deployment-ml/</guid><description>
&lt;h1&gt;모델 배포와 서비스 릴리스를 분리한 Uber의 ML 서빙 (2021-06)&lt;/h1&gt;&lt;div class="hx:overflow-x-auto hx:mt-6 hx:flex hx:rounded-lg hx:border hx:py-2 hx:ltr:pr-4 hx:rtl:pl-4 hx:contrast-more:border-current hx:contrast-more:dark:border-current hx:border-blue-200 hx:bg-blue-100 hx:text-blue-900 hx:dark:border-blue-200/30 hx:dark:bg-blue-900/30 hx:dark:text-blue-200"&gt;
&lt;div class="hx:ltr:pl-3 hx:ltr:pr-2 hx:rtl:pr-3 hx:rtl:pl-2"&gt;&lt;svg height=1.2em class="hx:inline-block hx:align-middle" xmlns="http://www.w3.org/2000/svg" fill="none" viewBox="0 0 24 24" stroke-width="2" stroke="currentColor" aria-hidden="true"&gt;&lt;path stroke-linecap="round" stroke-linejoin="round" d="M13 16h-1v-4h-1m1-4h.01M21 12a9 9 0 11-18 0 9 9 0 0118 0z"/&gt;&lt;/svg&gt;&lt;/div&gt;
&lt;div class="hx:w-full hx:min-w-0 hx:leading-7"&gt;
&lt;div class="hx:mt-6 hx:leading-7 hx:first:mt-0"&gt;&lt;p&gt;&lt;strong&gt;원문 정보&lt;/strong&gt;&lt;/p&gt;</description></item><item><title>남는 자원을 빌려 쓰는 우버의 Ray on Kubernetes</title><link>https://docs.makgol.com/external/uber/ml-platform/ubers-journey-to-ray-on-kubernetes-resource-management/</link><pubDate>Tue, 10 Nov 2026 00:00:00 +0900</pubDate><guid>https://docs.makgol.com/external/uber/ml-platform/ubers-journey-to-ray-on-kubernetes-resource-management/</guid><description>
&lt;h1&gt;남는 자원을 빌려 쓰는 우버의 Ray on Kubernetes (2025-04)&lt;/h1&gt;&lt;div class="hx:overflow-x-auto hx:mt-6 hx:flex hx:rounded-lg hx:border hx:py-2 hx:ltr:pr-4 hx:rtl:pl-4 hx:contrast-more:border-current hx:contrast-more:dark:border-current hx:border-blue-200 hx:bg-blue-100 hx:text-blue-900 hx:dark:border-blue-200/30 hx:dark:bg-blue-900/30 hx:dark:text-blue-200"&gt;
&lt;div class="hx:ltr:pl-3 hx:ltr:pr-2 hx:rtl:pr-3 hx:rtl:pl-2"&gt;&lt;svg height=1.2em class="hx:inline-block hx:align-middle" xmlns="http://www.w3.org/2000/svg" fill="none" viewBox="0 0 24 24" stroke-width="2" stroke="currentColor" aria-hidden="true"&gt;&lt;path stroke-linecap="round" stroke-linejoin="round" d="M13 16h-1v-4h-1m1-4h.01M21 12a9 9 0 11-18 0 9 9 0 0118 0z"/&gt;&lt;/svg&gt;&lt;/div&gt;
&lt;div class="hx:w-full hx:min-w-0 hx:leading-7"&gt;
&lt;div class="hx:mt-6 hx:leading-7 hx:first:mt-0"&gt;&lt;p&gt;&lt;strong&gt;원문 정보&lt;/strong&gt;&lt;/p&gt;</description></item><item><title>초당 800만 건의 예측을 감당하는 Uber 피처 로깅</title><link>https://docs.makgol.com/external/uber/ml-platform/taming-ml-firehose/</link><pubDate>Tue, 10 Nov 2026 00:00:00 +0900</pubDate><guid>https://docs.makgol.com/external/uber/ml-platform/taming-ml-firehose/</guid><description>
&lt;h1&gt;초당 800만 건의 예측을 감당하는 Uber 피처 로깅 (2026-09)&lt;/h1&gt;&lt;div class="hx:overflow-x-auto hx:mt-6 hx:flex hx:rounded-lg hx:border hx:py-2 hx:ltr:pr-4 hx:rtl:pl-4 hx:contrast-more:border-current hx:contrast-more:dark:border-current hx:border-blue-200 hx:bg-blue-100 hx:text-blue-900 hx:dark:border-blue-200/30 hx:dark:bg-blue-900/30 hx:dark:text-blue-200"&gt;
&lt;div class="hx:ltr:pl-3 hx:ltr:pr-2 hx:rtl:pr-3 hx:rtl:pl-2"&gt;&lt;svg height=1.2em class="hx:inline-block hx:align-middle" xmlns="http://www.w3.org/2000/svg" fill="none" viewBox="0 0 24 24" stroke-width="2" stroke="currentColor" aria-hidden="true"&gt;&lt;path stroke-linecap="round" stroke-linejoin="round" d="M13 16h-1v-4h-1m1-4h.01M21 12a9 9 0 11-18 0 9 9 0 0118 0z"/&gt;&lt;/svg&gt;&lt;/div&gt;
&lt;div class="hx:w-full hx:min-w-0 hx:leading-7"&gt;
&lt;div class="hx:mt-6 hx:leading-7 hx:first:mt-0"&gt;&lt;p&gt;&lt;strong&gt;원문 정보&lt;/strong&gt;&lt;/p&gt;</description></item></channel></rss>