What do you do once you have built the perfect model? In Dataiku, the answer is simple. This blog post features a demo from the 2021 Dataiku Product Days session on deployment and project delivery in Dataiku. This session was presented by Jacqueline Kuo, Senior Data Scientist at Dataiku, and François Sergot, Senior Product Manager at Dataiku. If you need a refresher on operationalization to start things off, check out this blog post to be ready for the demo.
Build a Project Ready for Production
This demo covers Dataiku features that will help you build a proper project that is ready to be moved to production. The goal of the project is to predict whether credit card transactions are fraudulent or not using known 2017 data to predict unknown 2018 data. This demo will first walk you through some Dataiku features to get your project production ready, including:
- Data quality checks using metrics and checks
- Automated flows using scenarios and running custom code within scenarios
Pushing to Production
The core focus of the demo answers this question: How can you seamlessly push your project to production with Dataiku? Two different scenarios are considered here:
- Deploying your model to an automation node be used for batch scoring
- Deploying your model using API nodes to be used for real time scoring