UrbanPulse Data Lab
Build, Govern & Scale with Microsoft Fabric
Goals
Five goals take you from raw streaming data to a multi-agent orchestrator and source-controlled workspace on Microsoft Fabric. Each goal maps to one or more modules; pick a goal pill below to focus the view, or keep All selected to see the full lab in order.
UrbanPulse is the fictional SaaS product you'll be building. It's a Smart City operations platform aimed at municipalities that run multiple public-service domains - hospitals, transit, public safety - and want one operational view across all of them.
Live patient vitals (heart rate, SpO2) and movement events (admit, transfer, discharge, imaging, surgery) from a regional hospital network. Reference data for hospitals, wards, and staff comes from an Azure SQL operational database.
Live train positions, speeds, and delay status across the Red, Blue, and Green metro lines. Route topology and regions come from an Azure Cosmos DB document store.
A shared Region concept ties hospitals and trains together so an operator can ask one question - "Region A has train delays AND rising hospital load. What should we do?" - and get a single, grounded answer.
Across the day, you'll ingest these data sources into one Fabric workspace, transform them into curated Silver and Gold tables, build reports in Power BI and a real-time dashboard, and apply AI through Fabric Ontologies, Data Agents, and an Azure AI Foundry orchestrator.
Sign in to Fabric with your lab account, confirm the lab capacity is assigned, and keep Module 0 open as your setup reference. You'll create your workspace and items in M1–M3.
Walk through the UrbanPulse multi-tenant architecture, OneLake explorer, capacity, RBAC, and the medallion model you'll build today.
Bring in file-based, relational, and NoSQL data using three Fabric ingest patterns - Upload a facility catalog Parquet, Mirror an Azure SQL Hospitals/Wards/Staff DB with zero-ETL, and Mirror a Cosmos DB for NoSQL database (regions and train routes) - all converging on one Bronze Lakehouse.
Stand up three Eventstreams routing live data from Event Hubs into a Fabric Eventhouse - hospital vitals, hospital movement, Metra trains.
Curate the Bronze tables into a Silver Delta table with a Spark notebook, then expose Gold-tier SQL views on the Lakehouse SQL endpoint for downstream BI and agents. Dataflow Gen2 and Materialized Lake Views are noted as alternatives.
Write KQL across hospital and transit tables. Pin live tiles to a Real-Time Intelligence Dashboard and watch the city update in front of you.
Build a Direct Lake semantic model on the Bronze Lakehouse and ship a 1-page Power BI report - no import refresh, no DirectQuery latency.
Build a semantic graph over your Lakehouse + Eventhouse data. Define Patients, Trains, and the cross-domain Region anchor that powers the agents you'll build next.
Build the Hospital Operations Fabric Data Agent on live KQL data and validate it in the Fabric agent canvas with natural-language questions.
Repeat the Module 8 pattern for Transit so you finish with a two-agent multi-domain catalog ready for orchestration.
Wire your Fabric Data Agents into an Azure AI Foundry connected-agent orchestrator for cross-domain reasoning. No code.
Connect your workspace to an Azure DevOps Git repository, commit the Fabric items to a branch, and update the workspace from Git.