HeroramenWhatsApp
All work

TymeX / AWS cloud

AWS online feature serving

The AWS serving side of a lending feature store, connecting offline feature pipelines with DynamoDB-backed online access.

AWSDynamoDBDatabricks Feature Store
SYSTEM OVERVIEW

Illustrative flow based on the public summary; not an internal implementation diagram.

  1. 01

    Shared definitions

    Feature definitions support both training and live decisions.

  2. 02

    Offline pipelines

    Databricks Feature Store supports offline feature processing.

  3. 03

    AWS online serving

    DynamoDB provides the online serving path.

01 / THE PROBLEM

What needed to change?

Lending decisions needed an online serving path using the same feature definitions as offline model training.

02 / THE APPROACH

How I approached it.

  • Built the lending feature store end to end, including its AWS online serving path on DynamoDB.
  • Connected offline pipelines on Databricks Feature Store with online feature availability.
  • Shared feature definitions across training and live decisions.

03 / THE OUTCOME

What changed?

Made shared feature definitions available for both offline training and live lending decisions through separate serving paths.