Serverless data ingestion at Abbott.

The challenge

Abbott's data pipeline ingested high volumes of records into a backend system through synchronous, monolithic APIs. As volumes grew, response times for bulk operations had crept past 10 seconds — long enough that downstream services timed out, and operations teams were waiting on jobs that should have been instant.

The Lambda codebase had grown to 2,000+ lines with SQL scattered through handler files, making changes risky and onboarding slow. Database connectivity used a single approach (psycopg2) which couldn't be optimised per-environment.

The approach

I led a refactor of the backend to a serverless, asynchronous architecture on AWS Lambda. Key design decisions:

• Async processing for bulk APIs. Heavy operations were moved off the request path into background workers, with status tracked via S3-based state monitoring.
• Centralised SQL. Queries lifted out of Lambda handlers into a dedicated query layer — easier to audit, easier to test.
• Dual database access layer. Built a switching layer supporting both Redshift Data API and psycopg2, enabling per-environment optimisation (faster cold-start in some, lower latency in others).
• Concurrent load testing. Built a framework simulating 50 concurrent users to validate throughput claims under realistic conditions.
• CI/CD with full observability. Jenkins pipelines, structured logging, and Allure-driven automation reporting integrated end-to-end.

The outcome

Beyond the headline numbers, the refactor unlocked a maintainable codebase that the broader team could onboard onto in days instead of weeks, plus a CI/CD setup that caught regressions before they reached production.

Tech stack

Engagement delivered while employed at Wipro. Abbott is a trademark of Abbott Laboratories; Wipro is a trademark of Wipro Limited. References here describe prior employment work and do not imply any current relationship between Abhishree Labs and Abbott or Wipro.