Two volumes · 2026 Part of the Spark 4.0 from Scratch series

The practitioner's guide to Databricks, platform and AI.

Two books that together cover everything an engineer actually needs to ship on Databricks in 2026, Unity Catalog, Lakeflow, ingestion, orchestration, CI/CD in Volume 3; Mosaic AI, Agent Bricks, MLflow 3, the Multi-Agent Supervisor, Lakebase in Volume 4. Read either standalone, or both for the complete picture.

Get the books Read sample chapters

37

Chapters across two volumes

~1600

Pages of practitioner depth

18×

Author of tech books

2026

Latest features covered

Cover of Databricks for Practitioners: The Production Lakehouse Playbook by Ritesh Modi
Vol 3
Cover of Databricks for Practitioners: The AI Lakehouse Playbook by Ritesh Modi
Vol 4
From $29.00

What's inside

Two volumes. The topics that actually decide whether you ship.

Every chapter is written for engineers who need to ship, not "an introduction to," but "here is the architecture, here is the code, here is what breaks at scale, here is what it costs."

Volume 3 · Ch 22–37 · ~800 pages

The Production Lakehouse Playbook

Databricks Platform & Data Engineering

Get Volume 3 →

Unity Catalog

The three-level namespace, ABAC, governed tags, lineage, audit, and how UC becomes the platform contract.

Read more →

Lakeflow SDP

Spark Declarative Pipelines (formerly DLT), declarative bronze/silver/gold, data quality expectations, change data capture.

Read more →

Lakeflow Jobs

Orchestration on Databricks, task dependencies, event-driven triggers, retries, and the patterns that scale past 100 jobs.

Read more →

Asset Bundles & CI/CD

Declarative Automation Bundles, CLI/SDK/Terraform, and GitHub Actions with OIDC, the production deployment story.

Read more →

Lakehouse Architecture

Delta Lake, Iceberg/UniForm, liquid clustering, predictive optimization, the storage substrate for everything above.

Read more →

Performance Tuning

Photon, partitioning, clustering, caching, and the failure modes the Spark UI alone won't reveal.

Read more →

Volume 4 · Ch 38–58 · ~800 pages

The AI Lakehouse Playbook

Databricks Platform & AI Engineering

Get Volume 4 →

RAG on Databricks

Mosaic AI Vector Search, chunking strategy, retrieval evaluation, and the patterns that actually scale beyond a demo.

Read more →

Agent Bricks

Classification, information extraction, auto-tuning, and the cost-quality frontier. The declarative path to production agents.

Read more →

Multi-Agent Systems

The Multi-Agent Supervisor, MCP, agent evaluation, tracing every hop, when single-agent is the right choice.

Read more →

MLflow 3

Experiments, UC Model Registry, aliases, traces, evaluation. The operating system for ML and GenAI workflows.

Read more →

Mosaic AI Vector Search

Delta-sync vs direct-access indexes, embedding choice, hybrid search, performance, and cost.

Read more →

Feature Store on UC

Point-in-time lookups, online serving from Lakebase, and avoiding training-serving skew.

Read more →

MLOps & Lakehouse Monitoring

Alias-driven promotion, champion/challenger, traffic splitting, and drift detection that actually fires when it should.

Read more →

Who this is for

Written for engineers who ship.

If you've already written PySpark and used Databricks notebooks but you're not sure how the same code becomes a production system, start with Volume 3. If you have Volume 3's platform fluency and need to build the AI layer on top, Volume 4. No "here's what Spark is." We focus on the patterns that hold up under load, under change, and under audit.

  • ✓ Data engineers shipping production pipelines on Databricks
  • ✓ Platform engineers building the governance layer
  • ✓ Solutions architects designing Unity Catalog and Lakeflow estates
  • ✓ Senior engineers migrating from legacy Hive Metastore or external orchestration
  • ✓ Engineering managers evaluating Databricks for an enterprise rollout
  • ✓ ML engineers shipping production GenAI systems on Databricks
  • ✓ AI engineers building RAG, agents, and multi-agent applications
  • ✓ Data engineers extending into AI and Mosaic AI workloads
  • ✓ Solutions architects designing Lakehouse + AI platforms
  • ✓ Engineering managers evaluating Databricks for AI initiatives
  • ✓ Practitioners preparing for Databricks ML Associate and Professional certifications
Ritesh Modi, Head of AI at MarketOnce · ex-Microsoft Principal Forward Deployed Engineer · 18× Author

About the author

Ritesh Modi

Head of AI at MarketOnce · ex-Microsoft Principal Forward Deployed Engineer · 18× Author

Head of AI at MarketOnce. Previously Principal Forward Deployed Engineer at Microsoft. Author of 18 technology books on AI, cloud, and infrastructure, translated, adopted by universities, and distributed by Microsoft Azure to 100,000+ enterprise users. Speaker at Microsoft BUILD, .NET Conf, and Global Azure. Creator of Microsoft's open-source RAG Experiment Accelerator and LLMOps PromptFlow templates.

These two volumes are the synthesis of what works, written for engineers who are tired of reading marketing material.

riteshmodi.com → LinkedIn → GitHub →

FAQ

Frequently asked questions

Why are there two books? Should I buy both?
The series splits cleanly. Volume 3 (The Production Lakehouse Playbook) is the platform foundation, Unity Catalog, Lakeflow ingestion and SDP pipelines, Lakeflow Jobs orchestration, Asset Bundles, CI/CD with OIDC, performance tuning. Volume 4 (The AI Lakehouse Playbook) is the AI build-out on top, Mosaic AI Vector Search and RAG, Agent Bricks, the Multi-Agent Supervisor, MLflow 3, Feature Store, MLOps, Lakehouse Monitoring, Lakebase. Read either standalone. Read both for the full picture, from raw Auto Loader ingestion to a multi-agent assistant deployed behind the AI Gateway.
Which volume should I start with?
If you're still figuring out Unity Catalog, Lakeflow pipelines, and how to deploy and orchestrate work on Databricks, start with Volume 3. If your platform is already in place and you're staring at the AI side (Mosaic AI Vector Search, Agent Bricks, MLflow 3, multi-agent systems), start with Volume 4. Volume 4 explicitly assumes Volume 3-level platform fluency, but it doesn't require having read Volume 3.
Are these books for beginners or experienced engineers?
Experienced. We assume you've written PySpark, you understand notebooks, you know what a DataFrame is. The series is for engineers and architects who want to go from prototypes to production systems on Databricks, UC governance, Lakeflow pipelines, RAG, agents, MLflow, model serving, the full operational picture.
Does this cover the 2026 platform features, Agent Bricks, the Multi-Agent Supervisor, Lakebase, MLflow 3?
Yes. Volume 4 is built around 2026 features: Agent Bricks (classification + information extraction), the Multi-Agent Supervisor with MCP integration, Lakebase (managed Postgres in the lakehouse), MLflow 3 with UC Model Registry and traces, the AI SQL function family (ai_query, ai_parse_document, ai_classify, ai_extract, ai_gen, ai_forecast, vector_search), and the AI Gateway. Volume 3 covers the platform features as of the 2026 cycle, Lakeflow renames, governed tags, predictive optimization, Liquid Clustering, OIDC federation.
Will the books go stale when Databricks releases new versions?
Kindle readers get free addendum updates for major Databricks releases, DBR LTS bumps, MLflow versions, Agent Bricks API changes, new AI SQL functions. The architectural lessons are durable; the platform-specific code gets refreshed.
How is this different from the Databricks official documentation?
The docs tell you what a feature does. These books tell you when to use it, what its failure modes are, what it costs, how it interacts with the rest of your stack, and how to operate it after the demo. The opinionated practitioner perspective is the value.
Do I need a Databricks workspace to follow along?
Yes, most code samples assume an active Databricks workspace. The free Community Edition is enough for many V3 chapters. Cloud-specific features (Vector Search, Model Serving, Agent Bricks, Lakebase) require a paid workspace; the books flag which sections need what.
What format is best?
Kindle for readers who want lifetime free updates as Databricks evolves. Paperback for readers who want a desk reference they don't lose tabs on. Hardcover for the long-life desk copy. All three are available on Amazon.
Who wrote the books?
Ritesh Modi, Head of AI at MarketOnce, previously Principal Forward Deployed Engineer at Microsoft. Author of 18 technology books including the Azure for Architects series and Solidity Programming Essentials. Creator of Microsoft's open-source RAG Experiment Accelerator. Speaker at Microsoft BUILD, .NET Conf, and Global Azure.

Two volumes

Get the series.

Volume 3 is the platform foundation. Volume 4 is the AI build-out. Read either standalone, or both for the complete picture.

Volume 3

The Production Lakehouse Playbook

Unity Catalog, Lakeflow, and the Databricks Data Intelligence Platform, the production playbook for engineers who already know Spark.

Buy on Amazon

$29.00 Kindle · $39.99 Paperback · 16 chapters

Volume 4

The AI Lakehouse Playbook

Mosaic AI, Agent Bricks, Lakebase, and the production Lakehouse, the 2026 field manual for shipping AI systems on Databricks.

Buy on Amazon

$32.00 Kindle · $39.99 Paperback · 21 chapters

Databricks for Practitioners · 2 volumes

From $29.00 on Kindle

Buy