Streams, lakes and oceans – working with on big data with Azure ML
summary
Azure ML Studio model training stays under a ~10 GB limit while deployed models serve as REST endpoints; a demo pipeline uses Event Hubs → Stream Analytics to compute per-cycle averages and standard deviations for 21 engine sensors, calls an Azure ML web service for a binary maintenance prediction, writes results to DocumentDB and shows a green/red dashboard. Retraining happens through a retrain API triggered from the UI, and production endpoints default to 20–200 concurrent requests.
AI-written summary · about
chapters (28)
- 0:00Intro
- 0:40Who is Barbara
- 1:29Agenda
- 2:23Aerospace predictive maintenance
- 3:51What is big data
- 5:36Volume of big data
- 6:21Traditional approach
- 8:00Data ocean
- 9:57Velocity
- 10:08Data served
- 11:00BI and analytics
- 12:27Streams
- 15:01ML vs Big Data
- 17:45Online Learning
- 18:12Big Data
- 20:49Problem description
- 23:20How does ML work
- 25:21What I love about Cortana
- 26:57Data sources
- 27:28Publish as API
- 27:54Architecture
- 38:21Azure ML Architecture
- 38:57Standard approach
- 47:02Retraining
- 48:44Setting up a web service
- 51:17Retaining data
- 53:35Scaling web services
- 56:09Questions
related talks
34mMaking Data Scientists Productive in Azure
48mStreamlining your Machine Learning projects with Azure Machine Learning
44mPractical deep dive into machine learning techniques and MLOps
43mA lap around Azure Machine Learning services
48mMachine Learning in Azure Databricks
58mDeep Dive into Azure Automated Machine Learning
01h 01mAI and ML imbedded in Azure Data Platform
37mPracticing Data Science in the cloud