Location via proxy:   [ UP ]  
[Report a bug]   [Manage cookies]                
SlideShare a Scribd company logo

Making Data Scientists Productive in Azure

•
•
V
Valdas Maksimavičius

Slides from my talk at Big Data Conference 2018 in Vilnius Doing data science today is far more difficult than it will be in the next 5-10 years. Sharing, collaborating on data science workflows in painful, pushing models into production is challenging. Let’s explore what Azure provides to ease Data Scientists’ pains. What tools and services can we choose based on a problem definition, skillset or infrastructure requirements? In this talk, you will learn about Azure Machine Learning Studio, Azure Databricks, Data Science Virtual Machines and Cognitive Services, with all the perks and limitations.

1 of 90
Download to read offline
MAKING DATA SCIENTISTS PRODUCTIVE IN AZURE Valdas Maksimavičius
Quiz Microsoft Machine Learning Server Machine Learning for .NET Azure Machine Learning Service Azure Machine Learning Studio Azure Databricks Data Science Virtual Machine SQL Server Machine Learning Services Azure Cognitive Services
Inspiration for the talk One thing about Microsoft  -  they have multiple ways to solve the same problem
Making Data Scientists Productive in Azure
Making Data Scientists Productive in Azure
So what do you mean by saying “Making Data Scientists Productive in Azure”?
Making Data Scientists Productive in Azure
Making Data Scientists Productive in Azure
Making Data Scientists Productive in Azure
6 Data Science stories
Tom • Full stack software developer • .Net, Node.js, Vue, React Scan faces to decide what advert to serve
Azure Cognitive Services What is it? Azure services with pre-built AI and ML models What can you do with it? Add intelligent features to your apps
Azure Cognitive Services - Overview • Vision (e.g. face / scene / object recognition, video analysis) • Speech (e.g. speaker recognition, speech-to-text) • Language (e.g. translations, phrase extraction, QnA maker) • Decision (e.g. content moderation, anomaly detection) • Search (e.g. Bing search)
Making Data Scientists Productive in Azure
Making Data Scientists Productive in Azure
“bald”: 0.17
Azure Cognitive Services - Summary Key benefits: • Minimal development effort • Easy integration via HTTP REST • Built-in support with other Azure services • Containers support
Azure Cognitive Services - Summary Key benefits: • Minimal development effort • Easy integration via HTTP REST • Built-in support with other Azure services • Containers support Considerations: • Limited customization allowed • Limited support for Non-English languages
ML.NET What is it? An open source and cross-platform ML framework What can you do with it? Create custom ML models using C# or F# without leaving the .NET ecosystem
Making Data Scientists Productive in Azure
Making Data Scientists Productive in Azure
ML.NET - Summary Key benefits: • Powers products like Microsoft Defender, Outlook, Bing, PowerBI • Seamlessly integrates ML into .NET apps • AutoML functionality • Leverage TensorFlow or ONNX Considerations: • Limited support for other ML libraries
Mark • Business Analyst • Basics of statistical analysis Create a sales lead list
Azure Machine Learning Studio What is it? Drag-and-drop visual interface for ML What can you do with it? Build, experiment, and deploy models using pre-configured algorithms
Making Data Scientists Productive in Azure
Making Data Scientists Productive in Azure
Making Data Scientists Productive in Azure
Deploy as web services • Batch execution • Request / Response
Making Data Scientists Productive in Azure
Azure Machine Learning Studio - Summary Key benefits: • Interactive visual interface • Built-in Jupyter Notebooks for data exploration • Direct deployment of trained models as web services • Built-in support for other Azure services
Azure Machine Learning Studio - Summary Key benefits: • Interactive visual interface • Built-in Jupyter Notebooks for data exploration • Direct deployment of trained models as web services • Built-in support for other Azure services Considerations: • Limited scalability (the maximum size of a training dataset is 10 GB) • Online only • Limited support for custom Python/R code
Lucy • Machine Learning Engineer • Python, Scikit-learn, Keras, TensorFlow Estimate damage (repair cost) in auto insurance
Making Data Scientists Productive in Azure
Azure Machine Learning Studio Service What is it? Managed cloud service for ML What can you do with it? Train, deploy and manage models in Azure using Python and CLI
Azure Machine Learning Service - Overview • Python SDK • Data preparation • Compute targets • Experiment tracking • Deployment targets
Azure Machine Learning Service - Compute Targets
Azure Machine Learning Service - Compute Targets • Your local computer
Azure Machine Learning Service - Compute Targets • Your local computer • Linux VM in Azure • Azure Batch AI Cluster • Azure Databricks • Azure Container Instance • Apache Spark for HDInsight
Azure Machine Learning Service - Compute Targets
Making Data Scientists Productive in Azure
Azure Machine Learning Service - Experiment Tracking
Azure Machine Learning Service - Experiment Tracking
Azure Machine Learning Service - Deployment Targets
Making Data Scientists Productive in Azure
Azure Machine Learning Service - Summary Key benefits: • Central management of scripts and run history • Run model training scripts locally, and then scale out to the cloud • Deployment and management of models to the cloud or edge devices • Start development locally (offline)
Azure Machine Learning Service - Summary Key benefits: • Central management of scripts and run history • Run model training scripts locally, and then scale out to the cloud • Deployment and management of models to the cloud or edge devices • Start development locally (offline) Considerations: • Python only • Requires some familiarity with the model management model
Bradley • Data Scientist / Engineer • Apache Spark / SQL / Python / Scala • Wants to spend more time outdoors than exploring beta version tools Build an enterprise data lake and data science environment
Azure Databricks What is it? Spark-based analytics platform What can you do with it? Build and deploy models and data workflows
Azure Databricks - Overview Collaborative Workspace • Notebooks • User access • Git integration
Making Data Scientists Productive in Azure
Azure Databricks - Overview Collaborative Workspace • Notebooks • User access • Git integration Databricks Runtime • Apache Spark • Rest APIs • Libraries
Making Data Scientists Productive in Azure
Azure Databricks - Overview Collaborative Workspace • Notebooks • User access • Git integration Deploy Jobs & Workflows • Job scheduler • Notifications & logs • Multi-stage pipelines Databricks Runtime • Apache Spark • Rest APIs • Libraries
Azure Databricks - Overview Collaborative Workspace • Notebooks • User access • Git integration Deploy Jobs & Workflows • Job scheduler • Notifications & logs • Multi-stage pipelines Databricks Runtime • Apache Spark • Rest APIs • Libraries Security • Single sign-on (SSO) • Access control list (ACL) • Secrets via Key Vault
Making Data Scientists Productive in Azure
Azure Databricks - Summary Key benefits: • Probably the most mature development environment for ML on the Azure platform • Nicely integrated with other Azure services
Azure Databricks - Summary Key benefits: • Probably the most mature development environment for ML on the Azure platform • Nicely integrated with other Azure services Considerations: • Online only
Joshua • Data Scientist • Research and development “I need a sandbox to learn and evaluate new tools”
Data Science Virtual Machine What is it? A virtual machine with pre-installed data science tools What can you do with it? Develop ML solutions in a pre-configured environment
Making Data Scientists Productive in Azure
Azure Data Science Virtual Machine - Summary Key benefits: • Probably the most complete development environment for ML on the Azure platform • Reduced time to install, manage, and troubleshoot data science tools and frameworks • Virtual machine options include highly scalable GPU images • A dedicated geospatial with ArcGIS distribution Considerations: • Online only • You need to take care of VM management
I asked for Cloud and they said NO
Rick • Specializes in R • Not allowed to push data to Azure Create personalized treatment based on individual health data
Microsoft Machine Learning Service Server What is it? Cross-platform standalone server for predictive analysis What can you do with it? Build and deploy models with R and Python
Microsoft Machine Learning Server - Overview • A new name for Microsoft R Server • Install on Windows / Linux / Hadoop cluster • Deploy models as web services packaged as container images • Satisfy security and compliance needs of any enterprise
Making Data Scientists Productive in Azure
Microsoft Machine Learning Server - Summary Key benefits: • Built on a legacy of Microsoft R Server and Revolution R Enterprise • Advanced security options • Deploy R and Python models as web services
Microsoft Machine Learning Server - Summary Key benefits: • Built on a legacy of Microsoft R Server and Revolution R Enterprise • Advanced security options • Deploy R and Python models as web services Considerations: • You need to deploy and manage Machine Learning Server in your enterprise
SQL Server Machine Learning Services What is it? A built-in SQL Server feature to support machine learning What can you do with it? Execute Python and R scripts with relational data
Making Data Scientists Productive in Azure
SQL Server Machine Learning Services - Summary Key benefits: • Run your scripts where the data resides and eliminate transfer of the data across the network to another server • Encapsulate predictive logic in a database function
SQL Server Machine Learning Services - Summary Key benefits: • Run your scripts where the data resides and eliminate transfer of the data across the network to another server • Encapsulate predictive logic in a database function Considerations: • Assumes a SQL Server database as the data tier for your application • Limited scalability • Long list of known issues
Quiz Azure Cognitive Services Machine Learning for .NET Azure Machine Learning Studio Azure Machine Learning Service Azure Databricks Data Science Virtual Machine Microsoft Machine Learning Server SQL Server Machine Learning Services
valdas.maksimavicius@cognizant.com linkedin.com/in/valdasm
Azure Machine Learning Service - Deployment Targets
Azure Machine Learning Service - Deployment Targets Native support: • Azure Container Instance • Azure Kubernetes Service • Azure IoT Edge Requires rework: • Linux VMs • Other cloud providers
Azure Machine Learning Service - Deployment Targets Native support: • Azure Container Instance • Azure Kubernetes Service • Azure IoT Edge
Making Data Scientists Productive in Azure
Making Data Scientists Productive in Azure
Azure Machine Learning Service - Overview • Python SDK • Data preparation • Compute targets • Experiment tracking • Deployment targets
Welcome to Vilnius Microsoft Data Platform Meetup Looking for enthusiasts to share their stories
Me in 2015 “With just a few clicks, you can have a Hadoop cluster up and running”
Me in 2018
Azure Cognitive Services - Similar services
žemėlapis su strėlukėm Atnaujink
Making Data Scientists Productive in Azure
Making Data Scientists Productive in Azure
Making Data Scientists Productive in Azure
valdas@maksimavicius.eu linkedin.com/in/valdasm valdas.blog
Welcome to Vilnius Data Platform Meetup • Making Data Scientists Productive in Azure by Valdas Maksimavičius • Building Churn Prediction Model Using Azure Databricks, Sklearn and MLflow by Tomas Lukas Komar • Snacks & Networking Agenda:

More Related Content

Making Data Scientists Productive in Azure

  • 1. MAKING DATA SCIENTISTS PRODUCTIVE IN AZURE Valdas Maksimavičius
  • 2. Quiz Microsoft Machine Learning Server Machine Learning for .NET Azure Machine Learning Service Azure Machine Learning Studio Azure Databricks Data Science Virtual Machine SQL Server Machine Learning Services Azure Cognitive Services
  • 3. Inspiration for the talk One thing about Microsoft  -  they have multiple ways to solve the same problem
  • 6. So what do you mean by saying “Making Data Scientists Productive in Azure”?
  • 10. 6 Data Science stories
  • 11. Tom • Full stack software developer • .Net, Node.js, Vue, React Scan faces to decide what advert to serve
  • 12. Azure Cognitive Services What is it? Azure services with pre-built AI and ML models What can you do with it? Add intelligent features to your apps
  • 13. Azure Cognitive Services - Overview • Vision (e.g. face / scene / object recognition, video analysis) • Speech (e.g. speaker recognition, speech-to-text) • Language (e.g. translations, phrase extraction, QnA maker) • Decision (e.g. content moderation, anomaly detection) • Search (e.g. Bing search)
  • 17. Azure Cognitive Services - Summary Key benefits: • Minimal development effort • Easy integration via HTTP REST • Built-in support with other Azure services • Containers support
  • 18. Azure Cognitive Services - Summary Key benefits: • Minimal development effort • Easy integration via HTTP REST • Built-in support with other Azure services • Containers support Considerations: • Limited customization allowed • Limited support for Non-English languages
  • 19. ML.NET What is it? An open source and cross-platform ML framework What can you do with it? Create custom ML models using C# or F# without leaving the .NET ecosystem
  • 22. ML.NET - Summary Key benefits: • Powers products like Microsoft Defender, Outlook, Bing, PowerBI • Seamlessly integrates ML into .NET apps • AutoML functionality • Leverage TensorFlow or ONNX Considerations: • Limited support for other ML libraries
  • 23. Mark • Business Analyst • Basics of statistical analysis Create a sales lead list
  • 24. Azure Machine Learning Studio What is it? Drag-and-drop visual interface for ML What can you do with it? Build, experiment, and deploy models using pre-configured algorithms
  • 28. Deploy as web services • Batch execution • Request / Response
  • 30. Azure Machine Learning Studio - Summary Key benefits: • Interactive visual interface • Built-in Jupyter Notebooks for data exploration • Direct deployment of trained models as web services • Built-in support for other Azure services
  • 31. Azure Machine Learning Studio - Summary Key benefits: • Interactive visual interface • Built-in Jupyter Notebooks for data exploration • Direct deployment of trained models as web services • Built-in support for other Azure services Considerations: • Limited scalability (the maximum size of a training dataset is 10 GB) • Online only • Limited support for custom Python/R code
  • 32. Lucy • Machine Learning Engineer • Python, Scikit-learn, Keras, TensorFlow Estimate damage (repair cost) in auto insurance
  • 34. Azure Machine Learning Studio Service What is it? Managed cloud service for ML What can you do with it? Train, deploy and manage models in Azure using Python and CLI
  • 35. Azure Machine Learning Service - Overview • Python SDK • Data preparation • Compute targets • Experiment tracking • Deployment targets
  • 36. Azure Machine Learning Service - Compute Targets
  • 37. Azure Machine Learning Service - Compute Targets • Your local computer
  • 38. Azure Machine Learning Service - Compute Targets • Your local computer • Linux VM in Azure • Azure Batch AI Cluster • Azure Databricks • Azure Container Instance • Apache Spark for HDInsight
  • 39. Azure Machine Learning Service - Compute Targets
  • 41. Azure Machine Learning Service - Experiment Tracking
  • 42. Azure Machine Learning Service - Experiment Tracking
  • 43. Azure Machine Learning Service - Deployment Targets
  • 45. Azure Machine Learning Service - Summary Key benefits: • Central management of scripts and run history • Run model training scripts locally, and then scale out to the cloud • Deployment and management of models to the cloud or edge devices • Start development locally (offline)
  • 46. Azure Machine Learning Service - Summary Key benefits: • Central management of scripts and run history • Run model training scripts locally, and then scale out to the cloud • Deployment and management of models to the cloud or edge devices • Start development locally (offline) Considerations: • Python only • Requires some familiarity with the model management model
  • 47. Bradley • Data Scientist / Engineer • Apache Spark / SQL / Python / Scala • Wants to spend more time outdoors than exploring beta version tools Build an enterprise data lake and data science environment
  • 48. Azure Databricks What is it? Spark-based analytics platform What can you do with it? Build and deploy models and data workflows
  • 49. Azure Databricks - Overview Collaborative Workspace • Notebooks • User access • Git integration
  • 51. Azure Databricks - Overview Collaborative Workspace • Notebooks • User access • Git integration Databricks Runtime • Apache Spark • Rest APIs • Libraries
  • 53. Azure Databricks - Overview Collaborative Workspace • Notebooks • User access • Git integration Deploy Jobs & Workflows • Job scheduler • Notifications & logs • Multi-stage pipelines Databricks Runtime • Apache Spark • Rest APIs • Libraries
  • 54. Azure Databricks - Overview Collaborative Workspace • Notebooks • User access • Git integration Deploy Jobs & Workflows • Job scheduler • Notifications & logs • Multi-stage pipelines Databricks Runtime • Apache Spark • Rest APIs • Libraries Security • Single sign-on (SSO) • Access control list (ACL) • Secrets via Key Vault
  • 56. Azure Databricks - Summary Key benefits: • Probably the most mature development environment for ML on the Azure platform • Nicely integrated with other Azure services
  • 57. Azure Databricks - Summary Key benefits: • Probably the most mature development environment for ML on the Azure platform • Nicely integrated with other Azure services Considerations: • Online only
  • 58. Joshua • Data Scientist • Research and development “I need a sandbox to learn and evaluate new tools”
  • 59. Data Science Virtual Machine What is it? A virtual machine with pre-installed data science tools What can you do with it? Develop ML solutions in a pre-configured environment
  • 61. Azure Data Science Virtual Machine - Summary Key benefits: • Probably the most complete development environment for ML on the Azure platform • Reduced time to install, manage, and troubleshoot data science tools and frameworks • Virtual machine options include highly scalable GPU images • A dedicated geospatial with ArcGIS distribution Considerations: • Online only • You need to take care of VM management
  • 62. I asked for Cloud and they said NO
  • 63. Rick • Specializes in R • Not allowed to push data to Azure Create personalized treatment based on individual health data
  • 64. Microsoft Machine Learning Service Server What is it? Cross-platform standalone server for predictive analysis What can you do with it? Build and deploy models with R and Python
  • 65. Microsoft Machine Learning Server - Overview • A new name for Microsoft R Server • Install on Windows / Linux / Hadoop cluster • Deploy models as web services packaged as container images • Satisfy security and compliance needs of any enterprise
  • 67. Microsoft Machine Learning Server - Summary Key benefits: • Built on a legacy of Microsoft R Server and Revolution R Enterprise • Advanced security options • Deploy R and Python models as web services
  • 68. Microsoft Machine Learning Server - Summary Key benefits: • Built on a legacy of Microsoft R Server and Revolution R Enterprise • Advanced security options • Deploy R and Python models as web services Considerations: • You need to deploy and manage Machine Learning Server in your enterprise
  • 69. SQL Server Machine Learning Services What is it? A built-in SQL Server feature to support machine learning What can you do with it? Execute Python and R scripts with relational data
  • 71. SQL Server Machine Learning Services - Summary Key benefits: • Run your scripts where the data resides and eliminate transfer of the data across the network to another server • Encapsulate predictive logic in a database function
  • 72. SQL Server Machine Learning Services - Summary Key benefits: • Run your scripts where the data resides and eliminate transfer of the data across the network to another server • Encapsulate predictive logic in a database function Considerations: • Assumes a SQL Server database as the data tier for your application • Limited scalability • Long list of known issues
  • 73. Quiz Azure Cognitive Services Machine Learning for .NET Azure Machine Learning Studio Azure Machine Learning Service Azure Databricks Data Science Virtual Machine Microsoft Machine Learning Server SQL Server Machine Learning Services
  • 75. Azure Machine Learning Service - Deployment Targets
  • 76. Azure Machine Learning Service - Deployment Targets Native support: • Azure Container Instance • Azure Kubernetes Service • Azure IoT Edge Requires rework: • Linux VMs • Other cloud providers
  • 77. Azure Machine Learning Service - Deployment Targets Native support: • Azure Container Instance • Azure Kubernetes Service • Azure IoT Edge
  • 80. Azure Machine Learning Service - Overview • Python SDK • Data preparation • Compute targets • Experiment tracking • Deployment targets
  • 81. Welcome to Vilnius Microsoft Data Platform Meetup Looking for enthusiasts to share their stories
  • 82. Me in 2015 “With just a few clicks, you can have a Hadoop cluster up and running”
  • 84. Azure Cognitive Services - Similar services
  • 90. Welcome to Vilnius Data Platform Meetup • Making Data Scientists Productive in Azure by Valdas Maksimavičius • Building Churn Prediction Model Using Azure Databricks, Sklearn and MLflow by Tomas Lukas Komar • Snacks & Networking Agenda: