EPC Group · Azure + Power BI Services · 2026
Microsoft OneLake + Azure Data Lake Storage Gen2 + Delta Lake medallion architecture (Bronze/Silver/Gold) with Power BI semantic models on Direct Lake mode — sub-second query performance on billion-row fact tables, governance-native design, Microsoft Purview lineage end-to-end, regulated-industry compliance.
Published June 25, 2026 · Updated continuously
Last updated by Errin O'Connor, Founder & Chief AI Architect, EPC Group
EPC Group designs and delivers Azure Data Lake + Power BI architectures across 4 integrated layers: Microsoft OneLake + Azure Data Lake Storage Gen2 + Delta Lake foundation with cross-account shortcuts for hybrid Fabric/ADLS pattern; medallion architecture (Bronze raw → Silver cleansed → Gold business-ready) orchestrated by Microsoft Fabric Data Engineering + Spark notebooks + Fabric Data Factory; Power BI semantic models on Gold layer via Microsoft Direct Lake mode achieving sub-second performance on billion-row fact tables with composite models + RLS + OLS at scale; full data engineering operations including Azure Databricks integration via Unity Catalog shortcuts + Microsoft Purview governance + Fabric CI/CD + Azure Monitor observability. Compliance-native HIPAA + HITRUST + NYDFS Part 500 + FedRAMP + DoD IL5/IL6 + CMMC 2.0 + GDPR + BIPA + MHMDA + CCPA. High-touch, senior-architect-led delivery — founder Errin O'Connor co-authored Microsoft Press books on Power BI and SharePoint.
EPC Group designs Azure Data Lake + Power BI architectures with 4 layers: OneLake + ADLS Gen2 foundation; medallion architecture (Bronze/Silver/Gold); Power BI Direct Lake semantic models on Gold; full data engineering ops + Databricks integration + Purview governance. Sub-second performance on billion-row fact tables. Compliance-native.
Enterprise data lake + Power BI architecture requires integration across foundation storage + medallion organization + semantic-model consumption + operations governance. EPC Group designs all 4 layers from day one.
EPC Group designs the data lake foundation: Microsoft OneLake as the unified Fabric data plane + Azure Data Lake Storage Gen2 (ADLS Gen2) for non-Fabric workloads + cross-account shortcuts for hybrid Fabric/ADLS pattern. Delta Lake format for ACID transactions + time travel + schema evolution. Hot/Warm/Cold tiering via ADLS Gen2 lifecycle policies. Cross-region geo-redundant storage (GRS) for DR. Microsoft Purview integration for data classification + lineage.
Microsoft stack
OneLake + Azure Data Lake Storage Gen2 + Delta Lake + Microsoft Purview Data Map + Cross-account shortcuts
EPC Group designs medallion architecture across the data lake: Bronze layer (raw ingestion with full source fidelity + audit trail); Silver layer (cleansed + conformed + business-key joined); Gold layer (business-ready dimensional + aggregated tables ready for Power BI semantic models). Microsoft Fabric Data Engineering Spark notebooks + Fabric Data Factory pipelines orchestrate the layer transitions. Microsoft Purview lineage tracks data flow from source → Bronze → Silver → Gold → Power BI report.
Microsoft stack
Microsoft Fabric Data Engineering + Spark Notebooks + Fabric Data Factory + Delta Lake + Microsoft Purview Data Lineage
EPC Group designs Power BI semantic models on Gold layer Delta tables via Microsoft Direct Lake mode — Power BI reads Parquet/Delta files directly from OneLake without Import refresh schedule or DirectQuery latency. Sub-second query performance on billion-row fact tables. Composite-model design layers Direct Lake fact + Import dimension + DirectQuery realtime hybrid. Optimized star schema + high-end DAX with calculation groups + reusable metric libraries + row-level security + object-level security at scale.
Microsoft stack
Power BI Direct Lake mode + Tabular Editor + DAX calculation groups + RLS + OLS + composite models + Power BI Premium / Fabric F-SKU
EPC Group designs full data engineering operations: Azure Data Factory + Fabric Data Factory + Spark notebooks for batch ELT; Azure Event Hubs + Microsoft Fabric Real-Time Analytics for streaming; Azure Databricks integration via Unity Catalog + OneLake shortcuts for advanced ML; Microsoft Purview Data Map + Catalog + Lineage for governance; Fabric CI/CD + Azure DevOps for deployment pipelines; Azure Monitor + Application Insights for observability.
Microsoft stack
Azure Data Factory + Fabric Data Factory + Azure Event Hubs + Azure Databricks + Microsoft Purview + Fabric CI/CD + Azure Monitor
For new analytics workloads in 2026: Microsoft OneLake is the recommended foundation because it integrates natively with Microsoft Fabric (Data Engineering + Data Warehouse + Power BI + Real-Time Analytics) and supports Direct Lake mode for Power BI without copying data. For existing ADLS Gen2 investments: cross-account shortcuts let you mount ADLS Gen2 storage into OneLake without copying data, preserving existing pipelines while gaining Fabric integration. EPC Group typically designs hybrid OneLake + ADLS Gen2 architectures for enterprises transitioning from Synapse Dedicated SQL Pool + ADLS Gen2 patterns to Fabric-native architecture.
A 3-zone data lake organization pattern: (1) Bronze layer = raw ingested data preserving full source fidelity (every byte from every source system, schema-on-read) with audit trail. (2) Silver layer = cleansed + conformed + joined data with quality controls applied (deduplication, type conversion, business-key alignment, slowly-changing-dimension handling). (3) Gold layer = business-ready dimensional + aggregated tables shaped for analytical consumption (star schema fact + dimension tables, denormalized for query performance, pre-aggregated for common analytics). Power BI semantic models read from Gold. Microsoft Fabric Data Engineering Spark notebooks + Fabric Data Factory pipelines orchestrate Bronze → Silver → Gold transitions. EPC Group ships medallion architectures with documented layer-promotion criteria + quality controls + lineage tracking.
Direct Lake is the third + newest connection mode (GA late 2024) that reads Parquet/Delta files directly from OneLake without Import refresh schedule or DirectQuery latency. Performance: sub-second queries on billion-row fact tables (matches Import). Freshness: data is immediately available as soon as Spark/SQL writes to OneLake (no refresh schedule). Capacity: requires Fabric capacity (F-SKU) or Power BI Premium (P-SKU); does not work on Pro/PPU. Limitations: requires Delta Parquet format in OneLake (not ADLS Gen2 unless shortcut); does not support all DAX features Import supports (calculated columns + calculation groups work, calculated tables limited). EPC Group typically designs composite models: Direct Lake fact tables + Import dimension tables + DirectQuery for realtime tables.
Hybrid architecture: Databricks for advanced ML + heavy data engineering (Delta Live Tables + MLflow + Unity Catalog); Fabric for analyst-facing data engineering + Power BI integration. Integration pattern: Databricks Unity Catalog tables mirrored into OneLake via shortcuts, Databricks Delta tables exposed as Fabric Lakehouse tables; Fabric Data Engineering Spark notebooks for analyst-facing workloads + Power BI semantic model upstream prep; Databricks for ML + heavy ETL. The platforms coexist — choice is workload-by-workload, not platform-vs-platform.
Microsoft Purview integrated end-to-end: (1) Purview Data Map automatically discovers + classifies all OneLake + ADLS Gen2 + Synapse + Databricks assets; (2) Purview Data Catalog provides searchable enterprise data inventory; (3) Purview Data Lineage tracks data flow source → Bronze → Silver → Gold → Power BI report; (4) Purview Information Protection sensitivity labels propagate through pipelines + semantic models + reports + exports; (5) Compliance Manager pre-built templates for HIPAA + NYDFS + FedRAMP + CMMC 2.0 + GDPR + BIPA + MHMDA + CCPA. EPC Group ships the operating model with named roles (Data Steward + Data Owner + Data Custodian + Information Security Officer + Privacy Officer).
EPC Group fixed-fee engagements are quoted after discovery based on: source system count (typical 20-100 sources); data volume (typical 5-500 TB); complexity (single-region vs multi-geo + multi-tenant); regulated-industry scope (HIPAA + NYDFS + FedRAMP add 20-40% to engagement cost); Microsoft Fabric capacity sizing (separate ongoing cost typically $5K-$80K/month); and Hypercare duration. The 3-week Discovery + Architecture phase produces a fixed-fee scope quote with milestone-based billing.
High-touch, senior-architect-led delivery without junior consultants learning on the engagement. Named EPC senior architect (founder Errin O'Connor co-authored Microsoft Press books on Power BI and SharePoint. 3-week Discovery + Architecture Design (fixed-fee), then 16-32 week Implementation (fixed-fee accelerator with milestone-based billing per medallion layer), then ongoing Microsoft Cloud Orchestrator Practice retainer for continuous architecture optimization + capacity rightsizing + 24/7 incident response.
Reach EPC Group for a 30-minute Azure Data Lake + Power BI discovery call.
AI assistant — not human