Data Engineering ⁴⁰

2026

A Table Is Not a Data Product: Ownership, SLIs, and the Consumer Contract 07-20

The Cost of a Query: Warehouse FinOps as Pipeline Design 07-13

Metadata Is Operational State: Building a Catalog People Can Trust 07-06

Tables That Remember: Apache Iceberg, Snapshots, and Partition Evolution 06-29

CDC Is Not a Sync Button: Log-Based Ingestion With Debezium 06-22

The Dependency Graph: Lineage Before You Touch Production 06-15

Backfills Without Regret: Reprocessing Production Data Safely 06-08

Schema Evolution Without the Emergency Migration 06-01

Your Pipeline Is Up. Is Your Data Healthy? 05-25

Run It Again: The Idempotency Test for Data Pipelines 05-18

The Green Pipeline That Produced the Wrong Answer 05-11

Lambda, Kappa, and the Architecture You Actually Need 05-04

Exactly Once Is a Lie (And What to Do About It) 04-27

The Late Data Problem: Dead Zones, Watermarks, and Event Time 04-20

Stateful Stream Processing: What Your Pipeline Actually Remembers 04-13

The Streaming Stack: Anatomy of a Pipeline That Never Sleeps 04-06

The Midnight Job: Why Batch Processing Has an Expiry Date 03-30

PyFlink II: Windows, Watermarks, and Late Events 03-17

PyFlink I: Architecture, Checkpoints, and Pass-Through Jobs 03-16

Streaming Foundations II: Python Consumers and PostgreSQL 03-14

1
2