Every number in this book was produced by running the platform it describes.
Building the Platform is the third book in the Enterprise Data Architecture for the AI-Native Era series. It implements the design package of Book 2 on an open-source reference stack that runs on one 16 GB workstation under Docker Compose, and builds the four tiers in order. Each of the eighteen chapters was run on the lab: the commands, their output and the numbers in the text come from those runs, and each chapter ends with seven acceptance tests that a reader runs with one command. Where the lab took a shortcut, the chapter says so; where the build found a flaw in the design, the correction is recorded rather than hidden.
What you build:
The lab and the platform base: identity, secrets, telemetry, and a generated retail datasetDATA: change data capture into Kafka, an Apache Iceberg lakehouse, Airflow orchestration, quality gates, a catalog with lineageINFORMATION: dbt Gold models, a Cube semantic layer, Superset dashboards and alerts, ClickHouse serving, embedding and data sharingKNOWLEDGE: features, training, serving and drift monitoring with MLflow and Feast; graphs and decision tables; retrieval-augmented answers with evaluationDECISION: a decision service, OR-Tools optimization and simulation, an AI agent with MCP tools and OPA guardrails, and a loop that learns from outcomesRunning it: SLOs and alerts, a CI gate, audit evidence, go-live gates, and Kubernetes overlays mapping every service to AWS, Google Cloud and AzureEvery component comes with its cloud-managed and commercial equivalents, and the appendices list every image, version, port and command.
For data engineers, platform and MLOps engineers, analytics engineers and architects who want to see the architecture run before they commit to it.
Book 1, The Four-Tier Model, sets out the architecture. Book 2, Analysis and Design, produces the design this book builds.