Data-intensive AI Training and Inferencing with Azure Blob Storage | BRK192

Discover how Azure Storage powers OpenAI and Microsoft AI model training and end-user applications through seamless access to exabytes of unstructured data. Explore how Azure’s AI ecosystem integrates with Blob Storage through advanced features and APIs that you can use in your own AI workloads today. Learn to leverage tools like blobfuse2, Blob Connector for PyTorch, and AzCopy to optimize your training, fine-tuning and inference pipelines for both performance and cost. To learn more, please check out these resources: * https://github.com/Azure/azure-storage-fuse * https://github.com/Azure/azure-storage-for-pytorch * https://aka.ms/OpenAIcasestudy 𝗦𝗽𝗲𝗮𝗸𝗲𝗿𝘀: * Kyle Knapp * Vishnu Charan Tumatin Jeyaprakash 𝗦𝗲𝘀𝘀𝗶𝗼𝗻 𝗜𝗻𝗳𝗼𝗿𝗺𝗮𝘁𝗶𝗼𝗻: This is one of many sessions from the Microsoft Build 2025 event. View even more sessions on-demand and learn about Microsoft Build at https://build.microsoft.com BRK192 | English (US) | Cloud Platform Related Sessions: BRK191 -- https://build.microsoft.com/sessions/BRK191?wt.mc_id=yt_PLlrxD0HtieHgukvOrEw3CqZuKtxiu_wnM BRK141 -- https://build.microsoft.com/sessions/BRK141?wt.mc_id=yt_PLlrxD0HtieHgukvOrEw3CqZuKtxiu_wnM BRK142 -- https://build.microsoft.com/sessions/BRK142?wt.mc_id=yt_PLlrxD0HtieHgukvOrEw3CqZuKtxiu_wnM #MSBuild, #CloudPlatform Chapters: 0:00 - Overview of Azure Storage for AI Workloads 00:01:16 - Scale and Use-case of Azure Storage 00:05:54 - Exploring Scale Accounts and AI Pipeline Optimization 00:14:07 - Benchmark Results 00:26:21 - Section Overview: Installation and Key Features 00:31:02 - Explanation of fast parallel transferring with BLOB I/O 00:36:53 - Running Code Snippets to Manipulate Data Set Using Python Libraries 00:41:05 - Session Recap and Closing Notes