Three Stunning AI Services on AWS | by Dmytro Khmelenko | Sep, 2022

And ideas on how to involve them in your next application

Photo by fauxels

Nowadays, machine learning is being adopted in many fields of software development. In the beginning, we were all impressed by how trained models recognize pictures. But that was only the starting point.

Then it went wide. There are a lot of products that actively use machine learning. The recent release of GitHub Copilot demonstrates how AI can help even developers. They receive intelligent code completion and code generation out of the box.

The best part is that we don’t need to hold a Ph.D. to start using machine learning like in the past. Many platforms offer well-designed solutions ready for use via API.

Let’s explore a few services from AWS and see how we can make your next application AI-powered.

When the enterprise operates on big chunks of data, they want to do something with those data. We can analyze those data to find patterns and insights, but also use them for predictions.

That is accurate for time series data. And AWS has a dedicated service to it. AWS Forecast was designed to work with various time series. The mandatory requirement for every entry is the presence of a timestamp.

1* BSAN1AoYIVOC7gPyv09eA
Possible forecasting domains

The service has a few predefined setups for training and forecasting. That can be retail sales, website traffic, or inventory planning. If you have specific requirements, it is possible to customize the domain.

Then we need to import our data, set up columns, and start training. We will receive a forecasting model in our domain when the training completes.

AWS Comprehend is a natural language processing service. It can recognize key phrases from the text, syntax, and sentiment. It uses pre-trained models on provided texts. Additionally, it can find out the language of the text and extract personally identifiable information.

This service requires no extra configurations and is ready for usage. With its simple API, we can easily integrate it into the application. How would you use it?

For example, if we need to make sure that the text doesn’t include any personal information. Or when we want to watch the reviews of the products and react immediately to negative ones.

The example from above shows how to use Comprehend API in Python SDK. The response will contain the sentiment summary and the scores.

For this particular case, the sentiment will be negative and the score towards it will be over 0.99. We can integrate this piece into the application. So when the restaurant receives a such bad review, we can notify a customer success team immediately.

Amazon made good progress toward image and video recognition. A few years ago they announced a service Rekognition. It has many features and customization capabilities.

The service can do labeling, face recognition, parsing texts from images, and much more. It uses a huge set of trained models to accomplish different tasks.

1*f RW1dBcKiVFm9rAct3K7w
Photo by Brooke Cagle on AWS Rekognition

We can use the AWS console to get an impression of how the service works. After uploading the picture we receive the results the next second. It is amazing how fast the processing completes.

Similar to AWS Comprehend, this service has a simple API. A few lines of code, and we get a response with the image analysis. This is how easily we can bring image recognition to any application.

The area of application of this service can be wide. In case, we need custom labeling, Rekognition gives an option to train your model. We have to provide a dataset, label images, and then start training. Over time, we get a custom model available on Rekognition.

Like many other cloud providers, AWS invests a lot in machine learning. Every year developers are happy to try out new services. Entrepreneurs spawn new ideas about applying AI to solve problems.

The services in this article are easy to start with. But they are not the only ones. There are many of them. Awake your curiosity and explore them!

News Credit

%d bloggers like this: