ProFinda: ML platform migration

Context

The resource optimization product already used Rails, React and PostgreSQL. Its machine learning insights layer needed a new MLOps setup without requiring the application itself to be rebuilt. The migration therefore had to preserve the product's data and integration contracts.

System and approach

ML workflows and feature engineering moved from AWS Airflow into GCP Vertex Pipelines. PostgreSQL remained the source of truth for features, and a defined API connected the insights service back to Rails. This kept the ML migration scoped to the workloads and services that needed to change.

• Preserve the Rails, React, and PostgreSQL product surface.
• Move AWS Airflow ML workflows into GCP Vertex Pipelines.
• Keep PostgreSQL as the feature source of truth.
• Return insights through a defined Rails to GCP API boundary.

Delivered scope

The work covers the ML pipeline migration, feature workflows and integration between the existing application and Vertex AI.

Technology

• Ruby on Rails
• React
• Python
• FastAPI
• GCP Vertex AI
• PostgreSQL