Skip to main content
About meProjects

Five reference projects
from twelve years of ETL focus.

From group-wide CRM migration to spatial data processing - in chronological order, with domain context instead of marketing.

DI² - Data Integration² Platform

03.2026 - ongoing

Side project · ETL generator that produces maintainable, readable and auditable SQL pipelines from metadata - plain SQL as output, runnable in any SQL engine.

★ Side project · ETL generator from metadata - self-built from spec to productionDI² in detail
Ownership
Side project
Web
di2.marcus-belz.de

Vision

The ETL process is derived from the source system's metadata - tables, columns, types, relationships. The output is versioned SQL, runnable on PostgreSQL and SQL Server.

Three principles

  • Maintainable - one procedure per SQL file, every code change visible as a Git diff
  • Readable - plain SQL; engineers and analysts understand it without ramp-up
  • Auditable - every run verifiable in SQL tables, three logging levels via SELECT

Data quality is the product, not a feature. Business rules live as WHERE clauses in the metadata; faulty records land in an error table instead of silently in the DWH.

Tech stack

PostgreSQL 17Next.js 16React 19TypeScriptAuth.js v5Keycloak 26Tailwindshadcn/uiDockerNginxLet's EncryptGitHubHetzner VPS

Methodology

AI-driven development with Claude Code: specialized skills for every workflow step (Requirements → Architecture → UX → Backend → Frontend → QA → Review → Deploy → Security audit), human-in-the-loop at every checkpoint. Four environments running in parallel on one VPS.

Full description, infrastructure, workflow

Geo-based web portals

since 09.2023

Architecture and development work for geo-based web portals - specification, database design and implementation, with a focus on geodata processing inside the database.

Sector
GIS / geodata
Focus
Geodata processing inside the database

Role & focus

  • Architect for database design and data management
  • Specification and development of the database logic
  • Integration of heterogeneous sources into a central Postgres platform
  • Ensuring data quality and historization

Typical tasks

The usual tasks of in-database geodata processing - spatial analysis, performance tuning and index strategies for geodata.

Database logic

Processing and analysis logic runs server-side in the database - stored functions and procedures, plain SQL as the engine. No offloading of geo logic into the application layer.

Tech stack

PostgreSQLPostGISPL/pgSQLSQL ServerPythonGitJiraConfluence

Industry 4.0 reporting for production process optimization

07.2021 - 03.2022

Optimizing process steps and material quality through historical reporting ("lessons learned") across all industrial plants of a globally operating industrial group.

Employer
adesso SE
Client industry
Steel industry
Team
> 8 people

Starting point

  • Control systems for melting furnaces, rolling mills and other industrial plants
  • Performance-critical access - the control systems must never be slowed down
  • Manual material analysis and manual reporting as the historical practice

Solution

Change Data Capture (CDC) against Oracle as a low-impact connection to the production systems. The core challenge was scale: ~500 source tables across 6+ plants had to be extracted and quality-assured consistently. Instead of hand-coding per plant, this ran on metadata-driven pipelines - table and column selection, staging generation and historization configured rather than programmed. At its core the same pattern DI² builds on today.

Scope

6+
Industrial plants + labs
~500
Tables (estimated)

Tech stack

CDCOracleSQL ServerT-SQLSSISData VaultQlikView

EAI platform for logistics and damage controlling

11.2016 - 12.2018

Transport management for vehicle delivery and damage management along the supply chain.

Employer
Daimler TSS GmbH
Client industry
Automotive industry
Team
> 10 people

Task

  • Supply chain controlling - optimizing transport capacities
  • Route controlling across all delivering markets
  • Damage management - identifying damage hotspots per carrier, vehicle, route and transport mode

Implementation

Connecting several inconsistent source systems to a central logistics DWH. Reconciling planned against actually driven routes - including vehicles transported in circles. ETL development with SSIS and T-SQL, orchestration via Integration Services, Qlik as the frontend.

Challenges

Source systems inconsistent with each other - same vehicle ID, three country codes, four status models. Plausibility checking as the central task before any aggregation.

Tech stack

SQL ServerT-SQLSSISQlik SenseETL-Generierung

Group-wide CRM migration from a document-based legacy system

11.2014 - 05.2016

Replacing 30 document-based silo solutions with a central, cloud-based CRM suite. Gradual roll-out across several European countries.

Employer
Infoman AG
Client industry
Sanitary/building supplier
Team
> 5 people

Task & implementation

  • Migration of all legacy data from 30 Lotus Notes applications via the CRM Online web service
  • Building a migration database with staging, validation, history, transformation and load stage
  • Data processing via stored procedures, orchestration via SSIS, web service access through a 3rd-party component
  • Delta migration via hash comparison (BK hash + non-BK hash) - initial load plus daily delta runs
  • Master data management for lookup lists, option sets, column mapping and error-handling configuration

Data quality framework

A complete, transparent error picture on three logging levels. Business rules, value ranges, mandatory fields, lookup consistency and type conversion as WHERE clauses in the metadata - no hard-coded error handling in the code.

Data volume

30
Source applications
150
Interfaces
100k+
XML files
3 TB
Migration volume
5000
Attributes per file (max.)
400+
Person-days

Tech stack

SQL ServerT-SQLSSISDynamics CRMAzureXMLLotus NotesStored ProceduresHash-Delta

Project order consistent with the hub order: DI² - side project, Geodata & PostGIS, Industry 4.0 & steel alloys, Automotive & logistics, Data migration & Microsoft Dynamics 365.

Contact

Get in touchLinkedInXING