New to Dataiku DSS? Try out our NEW Quick Start Programs today and get onboarded on the product in just one hour! Let's go

Building & Deploying Predictive Models as Webapp on Kubernetes - Watch on Demand

Community Manager
Community Manager
0 min read 5 2 619

Among the broad range of possibilities to deliver data products in Dataiku DSS, webapps are a great way to enable end users to interact with predictive models. In this session, @vhoudebine (Senior Data Scientist - AI Consulting at Dataiku) goes through the different possibilities and best practices to develop, scale, and reuse customized interactive web applications through Dataiku DSS. More specifically, he develops from scratch an interactive Credit Scoring web application and deploy it on a Kubernetes cluster for scalability.  

Community Manager
Community Manager

Thank you everyone who attended last week's event. Unfortunately we did not have enough time to answer all of your questions so @vhoudebine was kind enough to answer the remaining questions.

1) Is there a Github repo where the flask application is available?

I’ll post it on github soon

2) Will web apps work on the community edition?

Yes they do

3) Do we need to load the Model at each call to the Endpoint ?

No, the model is loaded in memory when starting the API service and doesn’t have to be reloaded each time

4)Does that mean if a lot of users (more than 50 for instance) require a prediction at the same time, then the Model Endpoint should be déployed to Kubernetes ?

If you want to ensure high availability of your Model Endpoint to handle lots of concurrent queries, you can deploy your API service on a Kubernetes cluster. Kubernetes has an internal load balancer and will handle the dispatch of the queries to the different pods of the cluster.

5) Can you point us to web apps that are currently on the internet that are using this approach. (Can you point us to a specific currently publicly used example?)

And the blog article is here


Let us know if you have any other questions by commenting below!



In looking at the site

today.  All Twitter accounts I put in report 78% human and 22% company.

Stuck Interface.jpg


Not clear why?