Measurable success in production.

Context

Greenfield development of a global 5-region Hub-and-Spoke Azure Data Platform with strict compliance requirements.

Our Contribution / Solution

• Multi-tenant 5-region Hub-and-Spoke architecture on Azure with regional Databricks workspaces and Unity Catalog
• Highly secure network topologies (Private Endpoints, Azure Firewall, Hub VNETs) for GDPR, USMCA, and China CSL/DSL
• Entire infrastructure deployed as standardized Terraform (IaC), orchestrated via Databricks Asset Bundles (DABs)
• AI agents for Jira management, repository screening during pull requests, and Gantt charts via secure MCP servers

Result

• Delivery of a fully standardized Terraform MVP in just 8 weeks
• Significant reduction in infrastructure costs through consolidation of isolated factory workspaces into regional shared hubs
• An absolutely compliance-secure foundation for all future global AI initiatives

Context

Stabilization and performance optimization of the central Microsoft Fabric data platform.

Our Contribution / Solution

• Full architectural responsibility for over 1,200 dbt models within a Data Mesh and Medallion architecture
• Design and development of a reusable Python library (lib_pegasus) for the complete automation of recurring data engineering tasks
• Seamless, context-rich logging mechanisms and dbt tests for improved failure diagnosis in production

Result

• Development of a bespoke Python library for the full automation of data engineering tasks
• Substantial increase in development velocity across the entire internal team through standard libraries
• Complete stabilization of mission-critical ETL processes and elimination of all runtime-critical performance bottlenecks under massive data volumes

Context

Decomposition of a complex legacy reporting infrastructure into a scalable Azure Lakehouse.

Our Contribution / Solution

• Highly scalable Data Lakehouse based on Azure SQL, Databricks, and Delta Lake
• Delta Live Tables (DLT) and Apache Kafka for zero-latency processing of incoming sensor data (REST, MongoDB, Blob Storage)
• Refactoring of historical SSIS pipelines and complex T-SQL logic into scalable PySpark and Spark SQL workflows
• AI-supported search platform (Azure OpenAI + Azure AI Search) based on structured metadata

Result

• Reduction of ETL data processing times by over 80% and lowering of operational costs by 30%
• Decrease in data processing time by 80% (from 2 hours down to 15–20 minutes)
• Reduction of operational cloud operating costs by 30%
• Increase in overall system scalability by 400%

Context

Development of a GCP-based Snowflake solution for automated ERP and Planisware data streams.

Our Contribution / Solution

• Scalable Snowflake Data Warehouse on the Google Cloud Platform (GCP) following Medallion architecture principles
• Highly efficient interface to the Planisware data source utilizing Denodo virtualization layers
• Robust data pipelines utilizing Apache Airflow and Snowpark Python (SCD and CDC logic)
• CI/CD infrastructure based on GitLab and Terraform, automated deployment on Kubernetes clusters (AKS)

Result

• Reduction of data loading times from 160 down to 10 minutes
• Shortening of ETL load times from a previous 160 minutes to merely 10 minutes
• Near real-time data provision for MicroStrategy reporting across over 500 dm retail stores

Context

Data processing automation for dynamic pricing within the European energy market.

Our Contribution / Solution

• Highly scalable and resilient ELT pipelines based on Snowflake and Azure
• dbt development model combined with Dagster for consistent transformations
• Fully automated generation of Data Lineage to increase comprehensibility for internal business analysts

Result

• Establishment of fully automated dbt Data Lineage and reduction of the data error rate by 15%
• Decrease of erroneous data records within the production pipelines by 15%
• Reduction of development time for new data pipelines by 50%

Context

IoT real-time data processing and time series optimization utilizing Prefect and Snowflake.

Our Contribution / Solution

• State-of-the-art, hybrid Data Lakehouse architecture with Snowflake accommodating structured and unstructured time series data
• Continuous data streams via Snowpipe directly into the data platform
• Bespoke ingestion scripts in Python employing Prefect and dbt to orchestrate parallel data processes (multithreading)

Result

• Significant acceleration in the acquisition of solar installation sensor data through multithreading data pipelines
• Substantial performance enhancement in processing complex, time-critical sensor data via multithreaded execution
• Stable real-time data streams enabling faultless ad-hoc analytics and yield monitoring