Industry Examples on CSA Loom¶
These examples show how to build real-world analytics solutions using CSA Loom workloads. Each example is fully ported from the existing csa-inabox example set — source code, infrastructure-as- code, sample data, and step-by-step documentation.
v1 — 8 ported examples¶
-
Retail E2E (already Loom-shaped)
Lakehouse + warehouse + semantic model + Direct-Lake-Shim refresh + Power BI report end-to-end.
-
Fabric Data Agent (already Loom-shaped)
Read-only Q&A pattern over a lakehouse with per-source few-shot examples + identity passthrough.
-
Financial Fraud Detection
Real-time transaction scoring + Activator alerts + analyst Data Agent. ML lifecycle on Databricks. Power BI fraud-trend dashboards.
-
Healthcare Clinical Analytics
HIPAA-scoped patient data lakehouse + clinical Power BI reports. Sensitivity-label propagation. RLS for clinician roles.
-
IoT Streaming
Manufacturing sensor data → ADX → Activator → predictive maintenance. Eventstream + Eventhouse parity.
-
Cybersecurity (MITRE ATT&CK)
Endpoint telemetry + KQL detection rules + threat-hunting Data Agent. Sentinel integration. Per-workspace SOC patterns.
-
Manufacturing IoT
CMMC L2-aligned defense-industrial-base sensor analytics in GCC-High. ITAR-eligible patterns.
-
GeoAnalytics
Spatial analytics for environmental + civil-engineering workloads. ADX geo functions. Loom Maps service (v2).
v1.1 — remaining 17 examples¶
The following examples ship in v1.1 (Q3-Q4 2026):
- AI Agents
- Casino Analytics (Tribal Government)
- Commerce Economic Analytics
- Data API Builder
- DOT Transportation Analytics
- EPA Environmental Analytics
- Interior Natural Resources
- ML Lifecycle (Loan Default)
- NASA Earth Science / API-First Multi-Model
- NOAA Climate & Ocean
- Retail Demand Forecasting
- Streaming
- Tribal Health
- USDA Agriculture
- USPS Postal Operations
See the existing csa-inabox Examples index for the original (non-ported) versions of all 25.
Port pattern¶
Each example follows the structure:
- What you'll build — diagram + components used
- Prerequisites — capacity SKU + boundary
- Step-by-step — Bicep → notebooks → semantic model → reports
- Forward migration — what happens when Fabric reaches your boundary
- Per-boundary notes — Commercial / GCC / GCC-High / IL5
- Cost estimate — sample $/month
- Source code — link to
examples/fiab-<name>/folder
Per PRP-14 for the build plan.
Source code¶
Per-example source folders at examples/fiab-<name>/: - infra/ — Bicep additions per workload - data/ — sample data generator + seed - notebooks/ — Databricks notebooks - dbt/ — dbt models - semantic-model/ — TMDL files - activator-rules/ — JSON rule definitions - data-agent/ — agent config + example queries - power-bi/ — Power BI Project (.pbip) - tests/ — E2E + validation
Voice + framing¶
Per [[writing-voice-no-customer-framing]] memory rule: all example pages use generic federal-mission or generic-industry framing. Never customer-specific. Customer success stories belong in private briefings, not in the public docs.
✅ "A federal financial services agency that monitors transaction streams typically wants to detect fraudulent activity in real-time."
❌ "The IRS uses CSA Loom to detect fraudulent tax returns."