Azure ML/AI Integration and monitoring | Serverless360
← Return To Home

Azure ML/AI Integration and Monitoring

Introduction

This blog will give you an overview of the session “ML/AI, Integration and Monitoring” at the Integrate 2020 Remote event by Steef-Jan Wiggers, Microsoft Azure MVP.

This blog will cover the introduction to ML/AI, Azure services used to implement AI, integration and monitoring using AI and the importance of data.

Journey towards data driven economy

Steef-Jan explained how humans started to use data starting from 4000BC when wheel was invented. It was followed by the invention of printing press in 1439. Internet came into picture in 1980’s when human started to use data. Cloud was invented in 2000s which broadened the scope of data. Now currently we are using ML/AI to deal with data.

Journey towards data driven economy

Data driven solutions

Custom software solutions make use of the data to differentiate between the business of customers. Data is also used in Internet of Things. The data can be used to decide on business solutions. Data will also be used in Health care products in future using AI and ML.

“Data is the heart of Solution to any problem”

Data driven solutions

Machine Learning

In traditional programming, we will provide the input and the rules to find out the output to the program and we will get the data which is the answer to our question which is provided as input.

In Machine learning, we provide both the answers and data to the system which in turn will frame the rules.

ML Integration and monitoring

 Azure Machine Learning

Azure Machine learning provides the tools needed by the developers and data scientists to solve their problem. Some of the tools are as follows.

  • Visual designer
  • Jupyter Notebooks
  • R scripts
  • Visual code extensions
  • Machine Learning CLI
  • Open source frameworks such as PyTorch, TensorFlow, and Scikit learn and many more

Smart Building Project

Steef-Jan briefed the Smart Building Project installed in his office where there are several IoT devices installed in different locations within the office. The IoT sensors will ingest the data to the gateways. The scheduled Azure function app in turn ingests the data to the telemetry. This is again handled by the stream analytics job. The output data from the stream analytics job can be visualized in the Power BI dashboards. The data is also stored in Azure Blob storage for further reference.

Smart Building Project

They import the data from the Azure Blob as CSV file and transform them using the Apply SQL Transformation Task to get the relevant data. 70% of the data is used for creating rules and the remaining 30% is used for testing purposes. For this transformation they use the Booted Decision trees regression using a parameter range of 5 for all parameters.

Considerations with Machine Learning

Since data is the important part of machine learning, it is required to make sure that we follow the privacy and regulations while manipulating the data.

“The amount of data used must also be considered in machine learning because unwanted data may result in unwanted computations which in turn reduces the performance.”

 We should also make sure that the use of Machine Learning in a solution to a problem solves any business use case.

Artificial Intelligence

Artificial intelligence is the capability of a machine to imitate intelligent human behaviour. Through AI, machines can analyse images, comprehend speech, interact in natural ways and make predictions using data.

Demo on Artificial intelligence

A demo was done on artificial intelligence where the API from different cloud providers were used. The demo included an app to which an image is being uploaded. It is validated by different cognitive service providers which in turn provide different scores. The different cloud providers used in the demo application are as follows.

  • Microsoft Cognitive Face API
  • AWS Recognition detect faces
  • Google Computer vision API

Steef-Jan has used some Nuget packages to consume these APIs. He uploaded his image in the App and used the above AI providers.

Microsoft Cognitive Face API and AWS Recognition provided the information about the age, gender etc based on the image uploaded. Google computer vision API also provided some basic information on the uploaded photo.

AI Integration and monitoring

Consideration with AI

Following are the considerations for AI

  • Workloads
  • Costs
  • Compliance (GDPR, US Privacy Act)
  • Value

Innovations

Steef-Jan sighted few projects which have won in the Microsoft competitions. Those projects were based on AI and ML. He also mentioned that “Artificial Intelligence Market is like to reach USD 390 billion by 2025”.

Innovations

Azure Integration Services

Some azure integration services like Logic Apps, Service Bus, API Management and Event grids can be used along with Machine Learning and AI to solve some business problems.

Azure Integration Services

Smart Solution Demo

A demonstration was done using Azure cognitive services along with Logic Apps. The current COVID19 scenario was handled in the demo. A Logic app is triggered whenever someone posts with #COVID19 in twitter. The post is then analysed by the cognitive service sentiment analysis which provides a sentiment score.

Azure ML/AI Integration and monitoring

Monitoring

ML/AI can be used in monitoring projects where there is a need to periodically observe and check the progress of something. Monitoring also helps in keeping the track of events or data.

Tollbooth Scenario

He explained a Tollbooth scenario, where ML/AI can be used. Following is the flow of data in the tollbooth scenario.

  • Images of vehicle are uploaded to Blob storage
  • An Event grid is triggered by the blob created event
  • This in turn triggers an Azure function which analyses the image using the Cognitive vision API.
  • The Event grid also triggers some Azure Functions which processes the data and stores them into Cosmos DB
  • Another function with cosmos DB trigger stores the processed image in a Blob storage
Tollbooth Scenario

Demo – License Plate

A demo was done using AI and ML to process the images of license plate. The flow of data here is as follows.

  • The images of license plates are uploaded to blob storage
  • An Event Grid is triggered by the blob created event
  • The Event Grid in turn triggers the Azure Functions which calls the cognitive service
  • The cognitive services return the license plate numbers as response
Demo – License Plate

Conclusion

The session was concluded with the recap. Monitoring is one of the key main aspects of Machine Learning. Machine learning can be used along with other Azure integration services to solve the business problems. Microsoft Azure offers some free AI and Machine Learning Services which can be used for learning. Steef-Jan concluded by saying that though Machine learning may seem like a fun but there is a big picture.

Azure integration services
Author: Ranjith Eswaran

Ranjith started his career at Kovai.co and works as a Junior Software Engineer. He lives with the passion - "Don't Give Up".