Amazon has been introducing many useful tools to make machine learning easier for developers around the world. Continuing this trend of making machine learning more accessible and comfortable, Amazon’s AWS launched a brand new approach in adding machine learning predictions to developer’s products and processes by directly integrating those predictions with their database.
“This announcement is a ball about making it easier for developers to add machine learning predictions to their products and their processes by integrating those predictions directly with their databases,” said Matt Wood, VP of artificial intelligence at AWS.
Problems with Adding Models
It won’t be long until all applications have machine learning and artificial intelligence used inside them. It can be challenging to incorporate the machine learning models into your databases, analytics and business intelligence reports. For example, for a simple bad review which states that someone doesn’t like the product, the production company can’t run sentiment analysis and get back at the reviewer to make things right. The data is available in their database, and there are ML services for that. Building the prediction pipelines to move data between models and applications is the problematic part.
Traditionally, developers have had to perform manual work to take these predictions to make them part of a more extensive application, process or analysis and this manual work is large and complicated. These complications include tedious application-level code development to copy data between different data stores and locations then transform data between formats before submitting data to the ML models and then get the results in your applications. This work is time-consuming and moving data in and out of data stores complicates governance and security.
How Amazon is Making Machine Learning More Accessible
AWS has a new approach to easily add machine learning to machine learning predictions with applications and business intelligence dashboards. AWS allows developers to combine tools like Amazon QuickSight, Aurora and Athena with SQL queries. These make it easier to add ML predictions without the need to build custom integrations, learn separate tools, write complex code, move data, and one can do it without any experience in machine learning. Developers can now access more variety of data than they already could without additional coding making the development process fast and more manageable.
Amazon’s Aurora is a MySQL compatible database that automatically takes data into the application to run a machine learning model assigned by a developer. Next, the developers can obtain additional datasets more easily from the company’s serverless system, Athena. The final step is QuickAight, which is Amazon’s data visualisation tool. The combination of these three tools provides a far more efficient approach to the development of the machine learning models.
For example, the lead scoring model in sales is where you pick the most likely targets to convert to sales. Using machine learning prediction you don’t have to write any additional code, that would adapt to different codes incompatible with each other aka the ‘glue code’. You just have to take the lead scoring model built by the data science team then deploy it to SageMaker and order all sales queues by priority based on a prediction from the model. Wood says, “Today, to do lead scoring you have to go off and wire up all these pieces together to be able to get the predictions into the application.” With this new capability, you can get there much faster.
Explaining further, “Now, as a developer, I can just say that I have this lead scoring model which is deployed in SageMaker, and all I have to do is write literally one SQL statement that I do all day long into Aurora, and I can start getting back that lead scoring information. And then I just display it in my application and away I go,”
Advantages of this approach
As pointed out by Amazon, this process makes the machine learning simple and more accessible to the developers all around the world. With broader access to a large number of datasets, the development part has become much faster, and no additional code sequences need to be called. Anyone who can know how to write SQL can use machine learning predictions in their applications without custom code and can focus on other parts of the process with more efficiency.
On the business intelligence front, a lot of Amazon customers fuss how it’s frustrating to build and manage prediction pipelines before getting predictions from a model. On average, these developers spend days writing application code to move data between models and applications; Amazon QuickSight can be used instead to visualise and report all ML predictions.
Enjoyed this story? Join our Telegram group. And be part of an engaging community.
Provide your comments below
Credit: Google News