The AI Lakehouse: Architecting Data Platforms for AI
Published by Self-published · Tech
Explores how to architect data lakehouse platforms specifically designed to serve AI and machine learning workloads.
Who this book is for
Data platform architects, data engineers, AI and ML engineers, and technical leaders and CDOs deciding how a lakehouse should serve both analytics and AI workloads.
Table of contents
- The Convergence of Data and AI
- Lakehouse Architecture Fundamentals
- Apache Iceberg for AI Workloads
- Open Table Formats Compared
- Vector Search in the Lakehouse
- Semantic Layers for AI Agents
- Feature Engineering on the Lakehouse
- RAG Pipelines on Lakehouse Data
- AI Training on Lakehouse Data
- Data Governance for AI
- Observability and Data Quality
- Reference Architectures
Excerpt
Every organization building AI systems today faces the same architectural question: where does the data live, and how do AI workloads get to it? For most companies, the answer is messy. Analytics data sits in a data warehouse. Raw files live in a data lake. Embeddings go into a vector database. Features get computed somewhere else and served from a feature store. Training datasets are exported, transformed, and versioned in yet another system. Each of these systems has its own security model, its own governance policies, and its own operations team.
