Blog

Notes from the lakehouse.

Deep engineering write-ups, product thinking, and ecosystem dispatches from the team building DXData on top of Apache Iceberg.

EngineeringFeatured

How we run 4.2B rows/day through a single Iceberg catalog

A walkthrough of the partition strategy, metadata layout, and compaction schedule that keeps our largest tenant queryable. Lessons from a year of tuning hidden partitioning in production.

Maya Okafor

Apr 8, 202611 min

Branching production tables: a 6-month postmortem

We gave every engineer a git-style branch on their production tables. Here is everything that broke, what we shipped to fix it, and what we would not build again.

Ren Takahashi

Mar 22, 20269 min

Why we built Nessie-style branching into DXData

Most catalogs treat branching as an afterthought. We treated it as a load-bearing primitive. Here is the product thesis behind that decision and where the model is heading.

Priya Venkatesan

Mar 4, 20267 min

SOC 2 Type II, one year later

A candid look at what the first year of continuous SOC 2 Type II compliance actually cost, what we automated, and the three controls that caused the most pain.

Daniel Hwang

Feb 18, 20268 min

Migrating from Snowflake without a rewrite

How a mid-market analytics team moved 340 pipelines and 1,200 dashboards off Snowflake in eleven weeks — without rewriting a single dbt model.

Sofia Ramos

Feb 2, 20266 min

Our contribution to Apache Iceberg v2.11

We upstreamed partition-level column statistics to Iceberg core. Here is what the feature does, how it lands in the spec, and why it matters for partition pruning.

Noah Abrams

Jan 20, 20267 min

Query federation: what finally made it fast

Federated queries across Iceberg, Postgres, and object stores should not be slow. We rebuilt our planner around three ideas that finally made the fast path consistently fast.

Kai Lindstrom

Jan 7, 202610 min

A dbt adapter that does not fight you

We shipped an official dbt adapter this quarter. The interesting part is not the adapter itself — it is the branch-aware materializations that let dbt runs happen inside an open branch.

Elena Vasquez

Dec 15, 20255 min

A fintech built a real-time fraud model on DXData in 9 weeks

Case study: how Lumen Pay moved their fraud scoring pipeline off a Kafka + Redshift stack and onto DXData with a single on-call rotation.

Sofia Ramos

Dec 2, 20255 min

RBAC is not enough: attribute-based access in practice

Role-based access control gets you 60% of the way to a real data governance model. Here is the attribute-based policy language we added to cover the rest.

Daniel Hwang

Nov 14, 20256 min

Newsletter

Stay current with the lakehouse ecosystem.

One email per month. New engineering deep-dives, product launches, and notable Iceberg community updates. No fluff.