NK.
All work

Case study 02 of 08CompleteOct 2025 – Dec 2025

Tier-1 European Telecom Fabric Migration PoC

Teradata to Fabric. Rebuilt as Bronze, Silver and Gold.

Teradata-to-Microsoft Fabric modernization PoC using metadata-driven ingestion, Bronze/Silver/Gold processing, CDC/SCD Type 2, Delta tables, and Lakehouse/Warehouse architecture.

Role
Migration engineer
Fabric ingestion, transformation, CDC/SCD2, validation
Stack
Microsoft Fabric
Data Factory, Lakehouse, Warehouse, PySpark, Delta
Scale
3 source systems
Roughly 18-24 daily incremental source tables
Outcome
Reusable migration path
Metadata-driven ingestion and reporting-ready serving layers
01

The brief. What it took, and what it changed.

Problem

Modernize legacy telecom data flows from Teradata/BTEQ and mixed source formats into Microsoft Fabric while preserving validation, retryability, and reporting readiness.

System

Fabric Data Factory pipelines and notebooks orchestrating source ingestion, landing, transformation, validation, Gold loading, and Warehouse serving across dual Lakehouse/single Warehouse architecture.

Impact

Supported 3 source systems and roughly 18-24 daily incremental tables; strongest scale and enterprise data-engineering credibility story.

02

How it works. In 4 stages.

    01

    Legacy extraction

    Teradata, BTEQ, and mixed source files move into reusable Fabric ingestion paths.

    02

    Medallion processing

    Bronze, Silver, and Gold layers preserve validation, retryability, and reporting readiness.

    03

    CDC and SCD2

    Hash comparison, validity windows, active flags, and deleted tracking keep incremental history consistent.

    04

    Warehouse serving

    Lakehouse and Warehouse layers prepare migrated data for downstream analytics consumption.

03

What shaped the build. And what it runs on.

    Used metadata-driven ingestion for reusable migration flows.

    Applied MD5 hash-based CDC, rec_start/rec_end validity windows, active/deleted tracking, schema correction, UTF-8 standardization, JSON flattening, and idempotent execution patterns.

Built with

  • Microsoft Fabric
  • Fabric Data Factory
  • Lakehouse
  • Warehouse
  • OneLake
  • PySpark
  • Spark SQL
  • SQL notebooks
  • Delta tables
04

Proof. What's public, and what's on its way.

References

  • Project linkPublic architecture or validation example to add after approval.Pending

Visual artifacts

  • Architecture diagramdiagramTo approve
  • Medallion flowdiagramTo approve
  • Sanitized validation screenshotsscreenshotTo approve

Private project summaryMetric

Supported 3 telecom source systems in a Microsoft Fabric migration PoC.

Source
Resume extraction and project inventory
Confidence
High confidence
Public note
Client identity remains anonymized as a Tier-1 European telecom provider unless permission is confirmed.

Private project summaryMetric

Supported roughly 18-24 daily incremental source tables using metadata-driven ingestion and Delta processing.

Source
Resume extraction and project inventory
Confidence
High confidence
Public note
Keep source-system names and private pipeline details out of the public version.
05

Tradeoffs. And what stays private.

The tradeoff

Detailed source-system names should stay out of the public case study unless client permission is confirmed.

Kept anonymous

Keep client anonymized as a Tier-1 European telecom provider unless permission is confirmed.

Next case studyBI Migration. Validated line by line, report by report.Read it