Skip to content
Discuss a project
  1. Kiz8
  2. Work
  3. Real-time CDC lakehouse

Real-time CDC lakehouse

Kiz8 built an Iceberg lakehouse that cut worst-case data freshness from about 32 to about 18 minutes with zero row loss across hundreds of millions of records.

Problem

Downstream teams worked from raw Postgres change-data dumps rather than a consistent, typed source.

Approach

Land raw events in bronze, clean and type them into 40 silver tables refreshed every 15 minutes, then build curated data marts for analytics, AI agents and machine learning.

Measured outcome

Worst-case freshness: ~32 → ~18 minutes; zero row loss

Tech stack

Lake/Storage
Apache IcebergR2 Data Catalog
Ingest & CDC
Postgres CDC
Transform
DuckDBCloudflare WorkersCloudflare Containers
Start a conversation

From raw data to better decisions.

Send one paragraph about the data, the decision it should support, or the hypothesis you need tested. We reply with what we would look at first.

Discuss a projectExplore our research ›