Dataiku Applications: A Boost to Collaboration and Reuse

During the 2021 Dataiku Product Days, VP of AI Solutions for Excelion Partners Ryan Moore walked us through a data science project to demonstrate the conversion of projects into reusable applications on Dataiku. This demo focused on converting a project into two different applications: one catered to business users and the other packaged for use by data scientists and other data teams within that organization. In this blog post, we will walk through how to create and leverage a Dataiku application catered to business users of an organization.

Increasing Accessibility and Reusability

Dataiku applications aim to make your work more accessible and reusable by other members of your organization. Business users want a friendly interface where they can upload their own data and receive ready-to-use insights or results as output. Dataiku applications enable just this: advanced data practitioners can package up Dataiku flows and serve them to end-users in a scalable way, without business users ever needing to see or understand the complexity behind the scenes.

Getting Your Project Ready

Experience the full demo to see how the project was prepared using Dataiku. With Ryan Moore, adopt the role of a data scientist working for a bank and build a project to predict which applicants are going to default on the loans that they've applied for.

Converting Your Project to a Reusable Application

With Dataiku applications, you can get your project into business user hands in a very short amount of time. To convert your project to a reusable application, Dataiku gives you two options.

1. Convert Your Project Into a Visual Application: Choose this option to create an application consumable by business users.
2. Convert Your Project Into an Application-as-Recipe: Choose this option if the application is to be used by data teams.

In this blog, we will focus only on the first option. By choosing the visual application option, Dataiku will create a template for you to design a user interface application to interact with your project. There are a number of predefined configurations in Dataiku based on which data sets you want to expose and how you want to allow users to interact with your project.
In most cases, creating this user interface is a quick process and does not require knowledge of code. Whereas if you wanted to create a custom web application, it would likely take hours or days to build this type of an interface, assuming you had a web development skill set already.

There are a lot of options you can define in your Dataiku application. Some of these include:

• Upload File Dataset: This comes out of the box from Dataiku and allows business users to upload new files easily.
• Edit Project Variables: This allows users to interact with your project in ways that you define.
• Run Scenario: This allows users to run a scenario in your Dataiku project, kicks off the scoring of the uploaded data, and runs the logic you have created.
• View Dashboard: This is the ability for the business users to look at the results after you have run a scenario and processed data. Users can also download a dataset of the scored results that they can integrate with other processes.

Creating an Application Instance

Now, when your business users want to consume this project and run it themselves, they can create an application instance. This refers to creating a duplicate Dataiku project that encapsulates all the logic of the application underneath the user interface you previously defined. Your business users will be able to interact and upload data to their own application without having to involve the data science team.