Architecture Overview

The platform has two sections:

• Analytical (Workbench)
• Production (Execution)

The Analytical layer is for analysts to test and validate their logic and get the required approvals and compliance checks. It connects to the existing historical data available in the Data Lake and provides a space for free-form analytics and workflow-driven analytics. The end goal of using the analytical layer is understanding the data and using it to finalize the decision logic as an artifact bundle.

The Production layer can be completely isolated from the Analytical layer and is meant to execute artifact bundles created by the Analytical layer. If realtime production is required, it can be run as a simple API or a Python function call. If batch production is required, it can be run as a Spark job.

Most of the discussion in this documentation concerns the Analytical layer. The Production layer setup will be like a connection to an existing production system, and hence depends on how that production system is currently set up.

In This Section

• Components: the services that make up the platform
• Networking: how those services talk to each other
• Scalability and Sizing: where to add resources when things slow down