Data warehousing & architecture
DMP — Data Platform Management
I bring data from all of your systems (CRM, ERP, website, files, API) into a single enterprise warehouse with order, governance and quality control — the foundation for BI, analytics and ML.
What is Data Management Platform
A Data Management Platform is a unified enterprise data platform (enterprise data warehouse/lakehouse) that collects data from all of a company's systems (CRM, ERP, web and mobile analytics, files, API) into one source of truth. I put it in order — cleansing, deduplication, a single business glossary — and make it available for reporting, analytics and models. In essence it's the foundation of enterprise analytics: BI and ML run on it, and cross-system dashboards and forecasts are built on top of it. This is about an enterprise data platform, not an ad-tech DMP with audiences and cookies.
What the service includes
Model and discovery
I survey the sources and design the target warehouse model — usually in the Data Vault 2.0 paradigm with layered storage and historisation, or Kimball/star for reporting.
Infrastructure and environments
I deploy the database/MPP, the orchestrator and secrets management, and separate the prod/dev/uat environments. The platform can be stood up on proven open-source in the client's infrastructure.
Sources and ETL/ELT
I connect sources and build the ingestion: incremental loads, change streams, S2T mappings. Source data lands in the warehouse without loss and then goes through transformations.
Data Quality
I set up scheduled data-quality checks, monitoring and alerts. Problems are visible before they reach reports and decisions.
Data governance
I maintain a business glossary, Data Lineage from source to report, a report registry and a role-based access model. Data becomes understandable and manageable.
BI, ML and support
I build dashboards and reporting, and where needed ML services (forecasts, monetisation). I provide training, documentation and SLA-based support.
What stages a DMP implementation consists of
- 01
Warehouse infrastructure
I deploy the core of the platform: an MPP database, the Apache Airflow orchestrator, secrets management in HashiCorp Vault, and separation of the prod/dev/uat environments.
- 02
Target model and S2T mappings
I design the layered warehouse model (staging → ODS → DDS → ADS) and describe the S2T mappings from source fields to the warehouse layers.
- 03
Metadata and business glossary
I populate the catalogue with metadata, maintain a business glossary and capture lineage. A shared language of terms and a map of data movement emerge.
- 04
ETL/ELT processes
I build and customise the ingestion and transformation processes: incremental loads, SCD2 historisation, layer-to-layer transitions on dbt (AutomateDV for Data Vault).
- 05
BI reporting (+optional ML)
I build ADS marts, dashboards and reporting for the business. Where needed, I add ML services on top of the marts — forecasts and data monetisation.
- 06
Operations and growth
I move into operations: monitoring, quality control, SLA-based support and the iterative onboarding of new domains and sources.
An example reference architecture of an enterprise data platform
The architecture is a vertical flow: sources → ingestion → warehouse with a layered model in the Data Vault 2.0 paradigm (staging → ODS → DDS → ADS) → consumption, with cross-cutting services running across all layers (orchestration, Data Quality, governance, masking). The stack is assembled from proven open-source components and deployed in the client's infrastructure without expensive licences.
Sources
CRM, ERP, web and mobile analytics, files and APIs — inconsistent, "dirty" data scattered across the company's systems.
Ingestion (EL / CDC)
I pull data into the warehouse: batch loads (Airbyte, dlt) or a change stream (Debezium + Kafka). Source data lands in the warehouse without loss.
Warehouse — DWH
The core with a layered model: Staging (raw, "as is") → ODS (cleansed, current) → DDS (detailed historised layer; record versions over time are captured on the SCD2 principle — in Data Vault these are satellites) → ADS (marts). Transitions between layers are built by dbt, and for Data Vault by AutomateDV.
Consumption
BI dashboards (Superset, Metabase) and ML services — predictive models and data monetisation — run on the ADS marts.
Cross-cutting services
Connected to all layers at once: orchestration (Airflow/Dagster), Data Quality (Soda/Great Expectations/dbt tests), governance (catalogue, lineage, glossary, role-based access) and masking for dev/test.
Projects
Analytics platform for the digital agency YA AGENCY
Analytical data warehouse and dashboards for Ash
End-to-end analytics and infrastructure migration for WiseAdvice-IT
Dashboards for the SaaS platform Entera
Analytical data warehouse for a chain of bars and restaurants
Analytical data warehouse for the Neolombard pawnshop chain
Unified data warehouse and dashboards for the delivery service Edostav
Data warehouse for the financial analytics project VISUALSECTORS
Tech stack
Apache Iceberg 
Airbyte 
dbt 
OpenMetadata Clients
Testimonials
I bring data from all of your corporate systems into a single reliable warehouse — one source of truth instead of fragmented, contradictory reporting. It’s the foundation of enterprise analytics: BI dashboards and ML models run on it, and cross-system reports and forecasts are built on top. I put the data in order, ensure its quality and manageability, and then provide stable operation and the further development of the platform.
Shall we discuss your task?
FAQ
How does an enterprise data platform (DMP) differ from MDM? +
MDM is a narrow layer about authoritative master records: customers, products, counterparties. An enterprise data platform (DMP) is a broad layer: the whole warehouse, all data domains, ETL/ELT and reporting. MDM usually lives inside or alongside such a platform, supplying it with reference data of consistent quality.
Is an expensive commercial product mandatory, or can it be built on open-source? +
A ready-made commercial product isn't mandatory. I assemble the platform from open-source modern data stack components — there usually isn't a single all-in-one product with a UI, but the proven components cover every layer. This makes it possible to deploy the platform without expensive licences, in the client's infrastructure.
What is Data Vault 2.0, and why the layered model and historisation (SCD2)? +
Data Vault 2.0 is an approach to warehouse design that's resilient to changes in the sources and convenient for adding domains. The layered model (staging → ODS → DDS → ADS) separates raw data, cleansed data, the detailed historised layer and the marts for reporting. SCD2 historisation stores every version of a record over time, so you can reconstruct the state of the data on any date and build correct analytics.
How long does implementation take? +
The timeline depends on the number of sources, their complexity and the state of the data. A basic platform circuit with the first marts usually launches in anywhere from a few weeks to a few months, after which domains are added iteratively. I prefer to move in stages: first a working core and clear value, then expansion.
How are data quality and governance ensured? +
I keep quality on scheduled checks (Soda, Great Expectations, dbt tests) with monitoring and alerts — problems are visible before they reach reports. I build governance on a catalogue with a business glossary, Data Lineage from source to report, and a role-based access model. As a result the data is understandable, traceable and manageable.
What about security, personal data (PII) and test environments? +
I segment access with a role-based model, store secrets in HashiCorp Vault, and keep the prod/dev/uat environments separate. For test environments I prepare de-identified copies: data masking and anonymisation, so that PII never reaches dev/test. That way development and testing run on realistic but safe data.