Service · Software development

Databricks specialist in the Netherlands.

We implement and migrate to the Databricks Lakehouse Platform, from Unity Catalog and Delta Lake to Databricks SQL, MLflow and Mosaic AI. Pragmatic, independent and free from vendor dogma. For Dutch enterprises, scale-ups and data teams who want to bring data, analytics and AI together on one platform.

Arrange an introductory call See our work

What exactly is Databricks.

Databricks is a unified data and AI platform, founded by the original creators of Apache Spark. Its core idea is the so-called Lakehouse architecture: a combination of a data lake (inexpensive, open storage for raw and structured data) and a data warehouse (fast, optimised SQL layer for analytics), with a complete machine learning and AI stack on top. Rather than copying data back and forth between a warehouse and an ML platform, everything runs on a single layer.

Under the bonnet sits Delta Lake, an open-source ACID transactional storage layer built on cloud storage such as S3, ADLS or GCS. On top of that, Databricks SQL with the Photon engine handles classic analytics workloads, Workflows orchestrate data pipelines, and MLflow manages model tracking, registry and serving. Unity Catalog takes care of governance, lineage and fine-grained access. Genie provides a natural-language interface to your data, Mosaic AI supports generative AI and fine-tuning, and Foundation Models host open models such as Llama and Mistral within your own account.

What our role as Databricks specialist Netherlands involves is translating that broad capability stack to your situation. We build lakehouse implementations greenfield, migrate existing warehouses, set up medallion pipelines and get Unity Catalog right from the start, before it becomes an expensive overhaul later.

What a Databricks project involves in practice.

Lakehouse
One platform for data, SQL, ML and AI
Multi-cloud
Runs on AWS, Azure and GCP, with EU regions available
Governance
Unity Catalog: lineage, audit and row-level security
Independent
No platform mark-up, and honest advice on fit

Three types of Databricks projects.

Depending on where your data landscape stands: greenfield, migration from a legacy warehouse, or optimisation and expansion of an existing Databricks environment.

Engagement 01

Greenfield lakehouse implementation

For organisations building a data platform from scratch

You have decided that Databricks will be the foundation for your data, analytics and AI work. We set up the workspace structure, configure Unity Catalog from day one, build the medallion architecture (bronze for raw data, silver for cleansed and conformed data, gold for business-ready datasets), and develop the first ingestion pipelines from your operational systems. This includes cluster policies, cost control, an RBAC model, and CI/CD for notebooks and jobs. A foundation you can keep building on for years without it creaking as soon as a second group of users arrives. For the broader strategic picture around this, see our wider approach to data integration consulting.

Unity CatalogDelta LakeMedallionCI/CD for data
Engagement 02

Migration from a legacy warehouse

For those coming from Teradata, Oracle, DB2 or Snowflake

You have an existing warehouse that is hitting its limits: licence costs are rising, ML workloads don't fit, or the system is end-of-life. We map the current landscape, decide which workloads go to Databricks SQL and which to Spark/Photon, and migrate step by step. That covers SQL dialect conversion, rewriting ETL jobs to Delta Live Tables or dbt-on-Databricks, and rerouting BI integrations to Power BI, Tableau or Looker. No big bang, but running in parallel until confidence has been built, with attention to data validation between the old and new environments.

Snowflake → DatabricksSQL conversiondbt-on-DatabricksBI re-routing
Engagement 03

ML, streaming and AI expansion

For teams looking to build out their Databricks environment

The foundation is in place, but the organisation wants more: streaming pipelines on Kafka or Kinesis with Structured Streaming, an ML workflow with MLflow for training, registry and serving, and generative AI use cases via Mosaic AI or Foundation Models. Or a Genie implementation so business users can ask questions in natural language on validated datasets. Often this also involves connecting to a real-time analytics platform, or we work alongside a data science specialist on the MLOps side. We build the extension and make sure it fits within the existing Unity Catalog governance boundaries.

Structured StreamingMLflowMosaic AIGenie

What you take away from a Databricks project.

A working lakehouse environment, plus the governance, documentation and runbooks to build on it independently as an organisation.

Lakehouse architecture

Workspace setup, Unity Catalog, medallion layers and cluster policies, fully documented.

Pipelines & models

Ingestion, transformation and data modelling live in production and staging.

Governance & security

Access model, audit logging, lineage and row-level security via Unity Catalog.

Cost controls

Photon tuning, autoscaling, cluster policies and cost reporting per team.

Managed service (optional)

Monitoring, ongoing development, and quarterly reviews of costs and data quality.

When a Databricks specialist makes the difference.

Four patterns in which Dutch organisations most often engage us for Databricks projects.

Platform unity

Unifying data, BI and ML

Your analytics team works on a warehouse, your data scientists work in a separate Python environment, and your AI team is experimenting somewhere else again. One lakehouse ends the reality of three copies of every table and gives you Unity Catalog as a central access model.

Legacy under pressure

Warehouse hitting limits on cost or ML

A Teradata, Oracle or legacy Snowflake environment is becoming too expensive or is limiting what the data team can do. Databricks offers a migration path where ML, streaming and SQL workloads come together on one platform with open formats.

Multi-cloud strategy

Staying cloud-independent

Your organisation does not want to be tied to a single hyperscaler. Databricks runs on AWS, Azure and GCP with largely the same capabilities, and Delta Lake is an open format that remains readable outside Databricks. A serious lock-in mitigation compared with pure SaaS warehouses. It fits well with our broader enterprise software approach.

AI ambition with a solid foundation

Generative AI on your own data

You want to deploy LLMs, RAG applications or fine-tuning on your own data, but without moving the data outside your cloud account. Mosaic AI and Foundation Models make this possible within Databricks, provided the data, governance and cost foundations are in good order.

Databricks SQL, Delta Lake and Unity Catalog — briefly explained.

Databricks SQL is the SQL warehouse layer within Databricks. It runs on Photon, Databricks' query engine written in C++, and is designed for BI and analytics workloads with fast response times. It integrates directly with Power BI, Tableau and Looker through dedicated connectors, and offers a serverless variant so clusters scale up and down automatically. For many organisations, this is the starting point for evaluating Databricks as an alternative to Snowflake or BigQuery, especially if ML workloads are also on the horizon.

Delta Lake is the open-source ACID storage layer that underpins everything. Data is stored in Parquet files on your own cloud storage (S3, ADLS, GCS), with a transaction log on top that provides consistency, time travel and schema evolution. Because the format is open, your data is not locked into a proprietary engine: other tools such as Apache Spark, Trino, DuckDB or Snowflake via Iceberg translation can also access it. An important difference from classic warehouses, where your data is only readable within the product.

Unity Catalog handles the access model, lineage and governance. One central catalogue across all workspaces, with fine-grained permissions down to column and row level, audit logging of every query, and automatic lineage between tables, notebooks and dashboards. For organisations with compliance requirements (financial services, healthcare, government), this is often the decisive reason to take Databricks seriously.

How a Databricks engagement works.

01Introduction→ 02Discovery→ 03Architecture→ 04Build & operate
First conversation

Introduction

Which systems, what ambition, what pain points. No obligation.

Workshops & interviews

Discovery

Sources, volumes, BI stack, ML ambitions, compliance context and cloud choice.

Document & decisions

Lakehouse architecture

Workspace setup, Unity Catalog model, medallion structure, cluster policies.

Sprints & ongoing

Build & operate

First working pipelines within a few sprints, then phased expansion plus operations.

Databricks versus Snowflake, Microsoft Fabric and BigQuery.

The question we get asked most often: Databricks or Snowflake? The short answer: Snowflake excels as a pure SQL warehouse for BI and analytics workloads, with an excellent user experience for analysts. Databricks excels as a unified platform where data engineering, SQL, machine learning and generative AI come together in one place. If your use case is purely warehousing and BI, and you have no ML or streaming ambitions, Snowflake is often the more pragmatic answer. Once ML, streaming or AI come into view, the balance tips towards Databricks.

Microsoft Fabric and Azure Synapse are a strong choice if your organisation runs entirely on the Microsoft stack: Power BI is central, identities sit in Entra ID, and the IT department has a Microsoft enterprise agreement. For pure Microsoft shops, that is an efficient route. Databricks is the choice if you specifically don't want to be tied to a single hyperscaler, or if the Lakehouse architecture and Spark engine suit your workloads better. Databricks also runs very well on Azure, so the choice doesn't have to be "against Microsoft".

BigQuery with Vertex AI on GCP and Redshift with SageMaker on AWS are the hyperscalers' own alternatives. Both work well within their own ecosystems, but they connect the warehouse and ML less tightly than Databricks does. Self-hosted Spark plus Delta Lake is an option for organisations with a strong in-house platform team that don't want to pay a managed-platform premium, though with considerably more operational burden. We advise independently on this trade-off; we don't earn a margin on Databricks licences.

The Databricks stack we work with.

The full breadth of the Databricks platform plus the surrounding tooling. We choose per project what fits, based on scope, cloud choice and existing stack.

Core Databricks
Unity CatalogDelta LakeDatabricks SQLWorkflowsDelta Live TablesPhoton
ML & AI
MLflowMosaic AIFoundation ModelsGenieFeature StoreModel Serving
Cloud & integration
AWSAzureGCPdbtKafka / KinesisPower BI / TableauTerraformPython / Scala / SQL

Frequently asked questions about Databricks in the Netherlands.

What is the difference between Databricks and Snowflake?
Snowflake is a pure SQL warehouse: exceptionally good at classic BI and analytics workloads, with a polished user experience for analysts. Databricks is a unified platform where SQL warehousing, data engineering, machine learning, streaming and generative AI come together on a single layer. For organisations that only do BI on a structured data model, Snowflake is often the more pragmatic answer. For organisations that want to add ML, streaming or AI use cases, or that don't want to maintain three copies of the same data across three systems, Databricks is usually the better fit. We advise independently on a case-by-case basis and earn no margin on licences for either platform.
How do Unity Catalog and governance work?
Unity Catalog is the governance layer within Databricks: a single central catalogue across all workspaces, with permissions down to column and row level, audit logging of every query, automatic data lineage between tables, notebooks and dashboards, and integration with cloud IAM. For organisations with compliance requirements, such as the financial sector, healthcare or government, this is often decisive. We set up Unity Catalog properly from the start; retrofitting it into an existing, ungoverned environment is more expensive and more painful than doing it from day one.
How do we keep Databricks costs under control?
Cost optimisation is a central part of almost every engagement. We work with cluster policies (which cluster types which teams may create), autoscaling on both job and interactive clusters, Photon tuning for SQL workloads, serverless where it proves cheaper, and periodic scans for long-running or idle clusters. We also provide cost reporting per team or project, so that the platform's users gain insight into their own consumption. For the licence costs themselves, we refer to the official Databricks pricing overview: that structure (per DBU and per workload type) sets the baseline, while the platform design determines whether you end up below or above it.
Can you help us migrate from Snowflake or a legacy warehouse?
Yes, this is a common engagement. We map the existing workloads, classify them (which become Databricks SQL, which become Spark jobs, which become Delta Live Tables or dbt models), set up the target architecture with Unity Catalog and medallion layers, and migrate in parallel. We run old and new side by side for a period, validate figures between both environments, and only cut over once confidence is established. For SQL conversion between Snowflake dialect and Spark SQL, we use tooling combined with manual review of the more complex queries.
Which Dutch or EU regions are available?
Databricks runs in all EU regions of AWS, Azure and GCP, including the Dutch Azure region and the EU regions of AWS and GCP in Frankfurt, Ireland and Paris. Data and compute remain within the chosen region, and with Unity Catalog plus customer-managed encryption keys you can demonstrate that the data does not leave your cloud account. For organisations with GDPR requirements and sector-specific compliance obligations, this can usually be covered. For large-scale processing of personal data, we help prepare the DPIA-relevant documentation on data flows and access model.
What AI and ML capabilities does Databricks offer?
A fairly broad stack. MLflow for experiment tracking, model registry and serving is integrated as standard. Feature Store for reusable features. Mosaic AI for generative AI workflows, including fine-tuning of open models. Foundation Models to host Llama, Mistral and Mosaic models within your own account. Genie as a natural-language interface to data for business users. And Lakebase (preview) as an operational database on top of Delta Lake. We help with use-case selection, as not everything that is technically possible is also worth building.
Do you work alongside our existing IT or data suppliers?
Almost always. We don't replace working parties where there's no need to. We often work alongside an existing BI supplier, your internal data team, or a cloud managed services partner. In that case our role is platform architect and builder on the Databricks side, while other parties keep theirs. For specific extensions, such as Python work on a dbt layer or streaming jobs, we sometimes bring in a Python developer. Knowledge transfer to your own team is standard: notebooks plus documentation, joint sprints and pair programming on critical components.
Do you combine advice with actual implementation?
Yes, that is a deliberate choice. We don't hand over a loose PowerPoint architecture for you to implement yourself. We write the architecture and build the pipelines, configure Unity Catalog, model the medallion layers and set up the first BI integrations. That keeps our advice sharp, because we are directly confronted with the assumptions we make. For organisations that prefer to build in-house, we can also act purely as an adviser and sounding board.

Talk to us about your Databricks project.

A no-obligation conversation of about half an hour. We listen to your situation (which data, what ambition, which cloud) and give direction. This includes cases where the outcome is that another platform or partner is a better fit.

Schedule an introductory callEmail fabian.vandijk@appfront.nl

Edit content