DevOps for AI in the Microsoft World
summary
Shows an end-to-end MLOps pipeline built with Azure DevOps + Azure Machine Learning Service and Azure Databricks that trains a Keras/TensorFlow encoder using GloVe embeddings on tens of thousands of car part descriptions, registers models as artifacts, and deploys to ACI or AKS. Pipeline dynamically provisions ML compute, runs a two-step train then evaluate flow, tags each model version with build number, enforces data snapshotting, and demonstrates a working AKS endpoint (model v6) with drift/telemetry hooks.
AI-written summary · about
chapters (18)
- 0:00Introduction
- 4:58Python SDK
- 9:09Data Versioning
- 16:20Challenges
- 19:24Security Integrity
- 22:53MLOps
- 25:28ML Pipeline Example
- 31:56ML Pipeline Requirements
- 32:32ML Pipeline Installation
- 33:28ML Cluster Setup
- 35:55Run Configuration
- 37:00Pipeline
- 39:37Train Evaluate
- 41:33Tensor Flow Model
- 43:08Model Fit
- 43:43Pipelines
- 45:00Compute Target
- 46:44Release
related talks
57mDevOps for Machine Learning
39mSimplify, speed and improve development with DevOps
24mFrom Model to Production: Automating the ML Lifecycle with CI/CD
56mDevOps for Machine Learning
45mMLOps and ML-Flops: going from dev to prod
32mProductionize Machine Learning on Azure
48mStreamlining your Machine Learning projects with Azure Machine Learning
42mCI/CD for Machine Learning