Building AI applications that leverage your data in object storage
summary
Azure Blob Storage is presented as the scalable, cost‑effective object store for training, fine‑tuning and RAG with concrete demos: mount containers with Blobfuse2 block caching to stream data and checkpoint a 3 GB model snapshot, run a 350‑node AKS benchmark over 2 PB using ~17,000 cores and hit 8.1 terabits/sec ingress (13.5 Tbps egress, ~36–37 TB/min read), integrate with Azure AI Search, Azure OpenAI, LangChain or Pinecone, and keep knowledge bases fresh and ordered via change feed. All features are GA and available to try.
AI-written summary · about
chapters (13)
- 0:00Introduction
- 2:01AI training workflows
- 6:37Storage challenges
- 8:43Azure storage solutions
- 13:16Blob fuse 2 demo
- 21:21Fine-tuning demo
- 23:49Scale testing
- 28:01Scaled accounts
- 29:17RAG architecture
- 33:06RAG demo
- 37:01Custom RAG setup
- 40:00Change feed feature
- 42:17Conclusion
related talks
55mAzure Cognitive Search Beyond the First Demo
50mBuilding AI ready applications
48mUsing Azure AI Foundry to manage all your Large Language Models
44mBuilding and scaling cloud-native intelligent applications on Azure
46mInside Azure innovations with Mark Russinovich
01h 01mStreams, lakes and oceans – working with on big data with Azure ML
37mHow to Accelerate Model Training and Eliminate the I/O bottleneck for the Cloud
31mAzure AI Innovations at Microsoft