Accelerating Data Science and Machine Learning Workflow with Microsoft Azure Machine Learning
Microsoft User Group Hyderabad AIML Day 2020
https://aditya-bhattacharya.net/
https://www.eventbrite.com/e/microsoft-user-group-hyderabad-aiml-day-2020-tickets-123940376001
MUGH
Accelerating Data Science and Machine Learning Workflow with Azure Machine Learning
1. Accelerating Data Science and
Machine Learning workflow with
Azure Machine Learning
~ Aditya Bhattacharya
Lead AI/ML Engineer,
West Pharmaceuticals Services
2. About Me
Currently, I am leading the Data Science team at West Pharmaceutical
Services and previously worked in Microsoft and well seasoned in domains
such as Machine Learning, Deep Learning, Internet of Things (IoT),
Robotics and Cloud Computing. In the Data Science domain, I am well
experienced with Computer Vision, Time-Series Forecasting, NLP, Speech
analysis.
Apart from my day job, I am an AI Researcher at an NGO called MUST
Research, and I am one of the faculty members at MUST Research
Academy. https://academy.must.co.in/
https://aditya-bhattacharya.net/
https://www.linkedin.com/in/aditya-bhattacharya-b59155b6/
3. Key topics for
discussion
1. Enterprise Data Science and Machine Learning
workflow
2. AzureML studio for 360○ workspace for
accelerating ML solutions from research to
production
3. AzureML Python SDK for building custom ML
pipelines
4. Smooth integration from other services in Azure for
a scalable ML solution
4. Industry 4.0
Expectations
• Data driven workflows and smart systems
powered by AI and Machine Learning
• Fill the gap between advance research and
leveraging benefits of technology at scale
• Minimize human intervention and human
bias and improve throughput using
Automated AI/ML solutions.
5. How does a scalable and
sustainable ML workflow look like?
6. Typical Enterprise MLWork Flow
User Interface Layer
Middleware
API Layer
Analytics LayerData Layer
Monitoring
Layer
Feedback
Layer
7. Typical challenges for Enterprise Data
Scientists for production level solutions
• Creating data pipelines for continuous flow of data
• Infrastructure needs -> need of flexible computing infrastructure to deal
with large volume of data
• Rapid prototyping ofAnalytical and Machine Learning models
• Model versioning and comparison
• Model monitoring and re-training
• Exposing model results as REST API endpoints for integrating with other
software applications
8. 360○
Solution
from Azure
ML Studio
• Integration with Jupyter Notebooks, AutoML and
designer view for rapid prototyping of ML models
and DS Solutions.
• Data registry, Model registry, Model deployment as a
Web service for accelerating from research to
production.
• Managing computes, data stores and data
manipulation tools under one umbrella.
12. Offline Model Flow
• OfflineTraining – Live Deployment
• For static data used for training – when the data is not dynamic and unknown to the training model
• Model Registering andVersioning through Azure ML
• Model Deployment Pipeline and result can be consumed through RESTAPI for real time inference
Offline Training Deploying trained model using Azure ML
Managed by user
application
13. Online Model Flow
• Online training – Live Deployment
• For dynamic data used for training – when the end user can configure the data to be
used for training the model – user configuration driven data
• Deployment on unknown samples and can be consumed in real-time
• Can be auto-tuned and can run without specialized attention by the Data Science team
Managed by user
application
Online training and deployment pipeline using Azure ML Managed by user
application
17. Summary
• Enterprise Data Science and Machine Learning
workflow
• AzureML studio for 360○ workspace for accelerating
ML solutions from research to production
• Building custom ML pipelines using Azure ML SDK
• Easy integration from other services in Azure for a
scalable ML solution
18. - ADITYA BHATTACHARYA
Questions?
-Want to connect over LinkedIn ?
- Or email me at: aditya.bhattacharya2016@gmail.com
- Follow me at: https://aditya-bhattacharya.net/
That’s all for today!