Practical deep dive into machine learning techniques and MLOps Q&A
summary
Live Q&A details how Azure Machine Learning handles real workloads: training and inference run on-prem and in cloud via Azure Arc hybrid compute, data import caches partitions for large datasets, and checkpointing recovers spot/failed runs. Federated learning appears as a preview integration with NVIDIA Flare, model monitoring uses a Model Data Collector that logs to blob storage, and a managed feature store provides immutable, versioned feature sets (ADLS Gen2 delta) plus registries and fine‑tuneable Hugging Face pipelines.
AI-written summary · about
chapters (13)
- 0:00Session introduction
- 2:25Hybrid ML and infrastructure
- 6:52MLOps and LLMOps
- 9:15Federated learning
- 12:03Data management and scale
- 14:31Monitoring and custom metrics
- 17:11Checkpointing and samples
- 19:29Feature stores and versioning
- 24:01Customization and model hub
- 28:28DevSecOps and integration
- 30:47Learning resources and data
- 37:22Azure ML vs Databricks
- 42:59Scaling with Registries
related talks
48mStreamlining your Machine Learning projects with Azure Machine Learning
27mBeginners guide to Azure MLOps
43mA lap around Azure Machine Learning services
58mDeep Dive into Azure Automated Machine Learning
21mMLOps for R with Azure Machine Learning
45mMLOps and ML-Flops: going from dev to prod
51mHow to get started with MLOps in Azure Databricks
41mTHE STATE OF MLOPS - MACHINE LEARNING IN PRODUCTION AT ENTERPRISE SCALE