Building observability without rewriting the application.

The problem

Logs and metrics existed, but observability was fragmented. Teams relied heavily on individual CloudWatch dashboards and queries, while some information was also being pushed into third-party tools. There wasn't a single place to understand what was happening across applications.

The idea

Instead of asking every application to change how it produced telemetry, I looked for a way to use the pipeline that already existed.

Why it mattered

The approach meant we could introduce a new observability layer without requiring application teams to rewrite their logging implementations.

Architecture

1. Application→
2. CloudWatch→
3. Subscription Filter→
4. Kinesis Data Firehose→
5. OpenSearch→
6. Unified dashboards