Be the first to read our first Ebook!

In the Age of AutoML, we are all Data Scientists! | Artificial Intelligence (AI) has become increasingly essential to our daily routine lives.

Create, share and run Data Science flowcharts with SmartPredict: The easiest way to complete a production-ready AI project

Published on May 31, 2021 by Haingomanitra H. F.

Completing a production-ready AI project requires high technical skills, both in terms of data science and software engineering. These make the use of AI expensive and time-consuming.  That's where SmartPredict comes in: a platform where users can complete their end-to-end AI project by simply dragging, dropping and configuring modules to build a data science flowchart that can be shared with others to be more collaborative. In other words, the easiest way to make an AI project successfully, with guaranteed time and cost savings. If you want to know how it all works, this article will tell you more about creating and sharing the data science flowchart in SmartPredict.

1-Instead of coding, create a data science flowchart

When we talk about performing an AI project, the first thing we think of is coding, installing packages, and working in different computing environments. That's the traditional process, but with SmartPredict, you'll carry out your AI project in a different way, the easiest, while being in full control of it. 

Well, in this platform, we have featured in configurable modules the comprehensive processes you probably need to build a data science flowchart by simply dragging and dropping into a workspace. Know that there are more than 100 modules in this platform, such as data visualizer, data preprocessing, feature engineering, Machine Learning (ML) and Deep Learning (DL) models, model training and evaluation, etc. All of them are based on open source frameworks and libraries such as Tensorflow, Scikit-learn, and many others.

For advanced users and code lovers who want to take control of their AI project, know that the module parameters are settable. That's not all, there is the Custom modules feature with which they can create their own module with Python code.

So far you know the principle of modules, but what about building the flowchart?

It is simply a matter of dragging and dropping modules from the module menus and interconnecting them to build it. There are two ways to do this: 

  • either with Manual Flow: that is, you build it from scratch manually ;
  • or with Autoflow: a process based on AutoML 2.0 with which the flowchart is generated automatically. 

Note that the flowchart is always customizable, whether it is built manually or generated automatically (module parameters are configurable, and customizable), giving you full control over your AI project.

For an end to end AI project with SmartPredict, you just have to build and launch 2 flowcharts: 

- the one in the Build space train and evaluate an AI model

- the other in the Deploy space is fully deployed as a web service to put the trained model in production.

To better understand how easy it is, let's take an overview with these two flowcharts: 

  • The flowchart in the Build space trains and evaluates the ML/DL model

It's just a typical DL ML workflow that trains and evaluates your model.

It is mainly composed of your dataset module, data preprocessing, your AI model, a model trainer, and evaluator. 

All these modules are included in the menu module, you just have to choose and configure it to get the desired process. Otherwise, you are free to customize one if you want to code specific dataset preprocessing or to build your own AI model.

Once built, you launch the flowchart and the value of the metric is displayed. By the way, it can be launched at will if you want to fine-tune parameters.

As an illustration, here is the flowchart to train and evaluate the Support Vector Classifier model for the iris classification project.

Flowchart to train and evaluate Support Vector Classifier model with the Iris Dataset:

  • The flowchart in the Deploy space is fully deployed as a Web Service to put the trained model in production

The deployment flowchart is also simple: it‘s typically constituted with a "Web service IN module", a data processing, the trained module, and the "Web service OUT module". 

And if you're not sure you can build it yet, there is an option that automatically generates it and can still be customized.

You just have to click on the rocket icon and it will be fully deployed as a web service. Once again, you can update it at will. And an API is generated automatically to call it in other environments. Otherwise, you can make tests and predictions in the Predict space

As an illustration, here is the deployment flowchart to deploy the trained model in the Iris classification project.

 

2-Sharing your flowchart for getting more collaboration with others

So far we just mentioned that the flowchart is customizable, but also know that it can be saved, shared, and used in other projects.

Nothing could be easier for this: just select your flowchart, save and download. You get a JSON file that you can share and upload for use in other AI projects in SmartPredict.

Not only can other teams collaborate on a project, but this saves you time if you are working on a similar project. Some modules may need to be modified and customized, you just run flowcharts and all is done.

The panel to save and share flowchart

Conclusion

Well, in this article, you have discovered the simplest way to run your AI project with SmartPredict. You don't need to code, just drag and drop modules to create a data science flowchart or generate it automatically with Autoflow. We also have also seen that users can take full of their AI projects as flowchart is customizable.  And for even more time saving and efficiency, flowcharts can be shared and used for other AI projects.

So what are you waiting for to succeed in your end-to-end AI project with SmartPredict? 

Try it in cloud.smartpredict.ai , it's free.

Thank you for reading. Long live and prosper everybody.

Please check these references to learn more about the concepts mentioned in this blog.