Breaking features

Exploit your data

Prototyping data pipelines and building AI apps that are fully deployable and scalable has never been easier.

Collaborative canvas

This collaborative environment allows you to share insights with your team and discuss solutions and tests in real-time. Thus, improving the quality of your results.

Button
marketplace

Using marketplace resources, you can select a model in seconds and test it with your data. You will be able to ship solutions faster and gather knowledge and insights from the community.

REST API & IO
Buttons

You can connect your PD canvas with your backend in seconds via the REST API. Also, there are many IO buttons available to trigger events when the canvas is in deploy mode.

Version control
LAB

Save your successful or compelling experiments for later review. Check it with your mates to find the best solution. You can always go back with the version control system.

Step-by-step guide

How does it work?

Make your first AI app in under 10 minutes. Use our Marketplace models to extract value from data.

Step 1

Start a new project

The first view of the Purple Dye platform shows the Dashboard. There you can see all your Canvases and its details, and you can access to the marketplace, to your profile, payment methods, etc. Is the place where to start creating a new Canvas.

Step 1

Start a new project

Step 2

The workspace

The Canvas is your AI app, where you can collaboratively build your data pipelines and connect all required buttons to make your app work. Built with UI/UX in mind, it can even be used on a small laptop with a supported web browser.

Step 2

The workspace

Step 3

Drag and drop

The first step in creating your Canvas is to drag and drop a button coming from the Marketplace. There you can find many buttons ranging from data preprocessing tools, Machine Learning models, Deep Learning models, connectors to external APIs and clouds, etc. One button after another, drag as many buttons as you require to build your AI-app.

Step 3

Drag and drop

Step 4

Connect

The buttons in the canvas, have to be connected so the datasets can flow across them. Just drag arrows from the edges, left to right to connect two buttons. You can test whatever you want, without limitations, without coding to iterate over your prototype and make it work.

Step 4

Connect

Step 5

Button configuration

Every button has its own configuration. It can be divided into two sections. The initial parameters that must be set to adapt the button to the incoming dataset, and the Mapper where you can connect the different data tensors as your app requires.

Step 5

Button configuration

Step 6

Train and test your experiments

Now is time to make our prototype work. You can run the entire canvas at once or you can do it dynamically one building block after the other. As the canvas can be run dynamically by blocks, you can train your models step-by-step without loosing the general plot. Everything runs distributed natively, so it scales to your demands.

Step 6

Train and test your experiments

Step 7

Reports and API

After you finish a run, you can check on the different outputs that the canvas provides. Configure the plots that you want to see, and wait for the execution. Finally, you can download a report with the overall results. Use those outputs and the canvas as your own data lab. Regarding to the output in terms of data, just connect the final data to a button of external API of a cloud provider and receive the results in your own infrastructure.

Step 7

Reports and API

Step 1

Library

Adapt your already existing Python 3 code with the Purple Dye library. Follow the API documentation steps to integrate your data processing algorithms into Purple Dye ecosystem.

Step 1

Library

Adapt your already existing Python 3 code with the Purple Dye library. Follow the API documentation steps to integrate your data processing algorithms into Purple Dye ecosystem.

Step 2

Github

Finally, you should adapt your model card further than writing the markdown file. Check that the results obtained with Purple Dye tests match your results, and upload a new button to Purple Dye marketplace.

Step 2

Github

Step 3

Button selection

After the bot verifies the scripts and detects the possible buttons to be added to the Marketplace, select those that you want to be shown publicly and those that you want to upload privately.

Step 3

Button selection

After the bot verifies the scripts and detects the possible buttons to be added to the Marketplace, select those that you want to be shown publicly and those that you want to upload privately.

Step 4

Model card

Step 4

Model card

Step 5

Facturation start

In the billing dashboard R&D can track their earnings. Also, researchers can open a paper claim process where Purple Dye staff will verify the authorship of the paper, so that the researcher can earn a commission of future marketplace buttons based on their paper.

Step 5