ExclusiveTechnical Guide

The open lakehouse architecture: Apache Iceberg, Polaris, and the semantic layer for BI and AI agents

Three teams, three systems, three different revenue numbers. This guide shows you how to build an Iceberg-native lakehouse with one copy of your data, a shared semantic layer, and open protocols your BI tools and AI agents can both trust.

What's Inside

How to price the real cost of data copies and spot the six surfaces of lock-in
What Parquet, Iceberg, Polaris, Arrow, and Ossie each do, and what they don't
A build sequence: federate first, define semantics, migrate progressively
How to give AI agents certified definitions through MCP and A2A

Chapters

Dremio

Exclusive Edition

By Alex Merced, Jean-Baptiste Onofré, and William Martin, PhD

Download the Technical Guide

Get instant, free access to the full guide.

GUIDE OVERVIEW

What You'll Learn

This guide walks the open lakehouse from the cost of copies to agent-ready semantics, across 17 chapters written for data engineers and platform teams.

Price the Problem of Copies

The first chapters count what copies cost, map the six surfaces where lock-in hides, and define what open has to mean before you evaluate a platform.

Build the Open Substrate

Parquet, Iceberg, Polaris, Arrow, and Ossie each get a chapter, including what each project doesn't do. Then follow the build sequence: federate first, define semantics, migrate progressively.

Prepare for AI Agents

Learn why agents need a semantic layer for definitions, grain, and permitted joins, and how MCP and A2A connect agents to data you've already governed.

Featured Contributors