Data Fabric in Action - How to enhance a Stock Trading App with ML and Data Virtualization

How do you query data across MongoDB, DB2, and flat files with a single SQL statement? See how a data fabric simplifies data access for machine learning.

#1about 2 minutes

What is a data fabric architecture?

A data fabric is an emerging architecture that integrates disparate data sources across hybrid multi-cloud environments to address distributed data challenges.

#2about 3 minutes

Common challenges in developing machine learning applications

Developers face significant hurdles in finding, understanding, integrating, and ensuring the quality of data before they can select and deploy ML models.

#3about 4 minutes

Exploring the components of the IBM Data Fabric

The platform architecture is built on four pillars—collect, organize, analyze, and infuse—and includes key automated services like a data catalog, data virtualization, and privacy controls.

#4about 3 minutes

Understanding roles and responsibilities in the AI lifecycle

A successful AI project involves collaboration between distinct roles like data engineers, data stewards, data scientists, and developers, each with specific tasks.

#5about 2 minutes

The platform architecture of IBM Cloud Pak for Data

IBM Cloud Pak for Data is built on Red Hat OpenShift, a Kubernetes-based platform that provides scalability, automated deployments, and a control plane for integrated services.

#6about 5 minutes

Use case: Enhancing a stock trading app with ML

To reduce customer churn, a stock trading application is enhanced with an ML model to predict churn risk and data virtualization to simplify access to diverse data sources.

#7about 5 minutes

Demo: Discovering data with the knowledge catalog

The platform's central catalog allows developers to search for data assets, view previews, and understand data context through automatically assigned data classes and business terms.

#8about 6 minutes

Demo: Building an ML model with the AutoAI experiment

The AutoAI experiment automates model selection by testing multiple algorithms and hyperparameters, presenting a leaderboard to help developers choose and deploy the best model as a REST API.

#9about 8 minutes

Demo: Setting up real-time data virtualization

Data virtualization allows connecting to heterogeneous sources like MongoDB and relational databases, presenting them as standard SQL tables that can be visually joined and queried in real time.

#10about 6 minutes

Q&A: Sharing assets and training data limits

The discussion covers how users can publish their own refined data assets to the catalog, considerations for training data size, and API connectivity for native mobile applications.