Skip to content
Code by Pawpu

2021–2023

ETL Data Processing System

Serverless AWS pipeline processing high volumes with auto-scaling Step Functions workflows.

Case study · Software Engineer · serverless data pipeline

Written by Al Beltran (Al Andrew Paul Beltran), Software Engineering Lead.

Desktop, tablet, and iPhone 16

01 · The problem

Fixed server pools struggle with bursty ETL workloads. Teams needed elastic throughput, monitoring suitable for SLAs, and APIs for downstream consumers.

A production-grade serverless ETL system on AWS that processes large volumes of records with orchestrated workflows and downstream APIs.

02 · My role

Software Engineer · serverless data pipeline

03 · The result

Built high-volume processing with Lambda, S3, and Step Functions; multi-stage transforms with production monitoring; API Gateway endpoints for internal and downstream consumers.

04 · Architecture

  1. 01

    S3 as landing and intermediate storage

  2. 02

    Lambda for transform and processing stages

  3. 03

    Step Functions for workflow orchestration

  4. 04

    API Gateway for downstream consumers

Implementation

  • High-volume serverless processing
  • Multi-stage transforms with monitoring
  • API Gateway endpoints for consumers
  • Elastic throughput without fixed server pools

Challenges & outcomes

Challenges

  • Designing multi-stage transforms that fail safely
  • Observability for production SLAs on serverless paths

Outcomes

  • Elastic throughput without managing fixed server pools
  • Production-grade orchestration suitable for daily high-volume loads

Lessons

  • Serverless ETL succeeds when orchestration and retries are first-class
  • Downstream APIs should not expose raw pipeline internals

Related writing