Winparts sells car parts to consumers in the Netherlands and across Europe. Their core promise: finding the right part for your car. That sounds simple, but is technically complex. The product range is enormous, fitment data is demanding, and every mistake leads directly to a return. Returns cost money and erode trust.

As a Dutch company transitioning into a European player, they needed a foundation that could grow with them. Not off-the-shelf enterprise tools, but a scalable platform built on open source: dlt for data ingestion, dbt for data models, and cloud-native orchestration. Set up cloud-agnostically, so the choice of GCP, AWS, Azure, or any other platform never becomes a constraint.

A data platform was only the beginning. The question many organizations will recognize: how do you add AI to this without building a second architecture?

Winparts' answer: AI jobs as a natural extension of the existing platform. Lightweight containers that read data from the warehouse, execute an LLM task, and write the result straight back. No separate tooling, no new infrastructure. The framework is set up once; after that you add use cases one by one.

The first use case: the returns process. Annotation at scale, pattern recognition, structural improvement points for assortiment and fitment data.

What you'll take away A hands-on session for anyone who wants to implement AI in an existing data platform — or hasn't started yet.

How to set up a data platform that is ready for AI, without starting over
How AI jobs work as a lightweight extension on an existing platform, cloud-agnostically
How to start with one concrete use case and scale from there
Honest trade-offs: which open source tools, which trade-offs, what it has concretely delivered

Winparts sells car parts to consumers in the Netherlands and across Europe. Their core promise: finding the right part for your car. That sounds simple, but is technically complex. The product range is enormous, fitment data is demanding, and every mistake leads directly to a return. Returns cost money and erode trust.

As a Dutch company transitioning into a European player, they needed a foundation that could grow with them. Not off-the-shelf enterprise tools, but a scalable platform built on open source: dlt for data ingestion, dbt for data models, and cloud-native orchestration. Set up cloud-agnostically, so the choice of GCP, AWS, Azure, or any other platform never becomes a constraint.

A data platform was only the beginning. The question many organizations will recognize: how do you add AI to this without building a second architecture?

Winparts' answer: AI jobs as a natural extension of the existing platform. Lightweight containers that read data from the warehouse, execute an LLM task, and write the result straight back. No separate tooling, no new infrastructure. The framework is set up once; after that you add use cases one by one.

The first use case: the returns process. Annotation at scale, pattern recognition, structural improvement points for assortiment and fitment data.

What you'll take away A hands-on session for anyone who wants to implement AI in an existing data platform — or hasn't started yet.

How to set up a data platform that is ready for AI, without starting over
How AI jobs work as a lightweight extension on an existing platform, cloud-agnostically
How to start with one concrete use case and scale from there
Honest trade-offs: which open source tools, which trade-offs, what it has concretely delivered

Winparts sells car parts to consumers in the Netherlands and across Europe. Their core promise: finding the right part for your car. That sounds simple, but is technically complex. The product range is enormous, fitment data is demanding, and every mistake leads directly to a return. Returns cost money and erode trust.

As a Dutch company transitioning into a European player, they needed a foundation that could grow with them. Not off-the-shelf enterprise tools, but a scalable platform built on open source: dlt for data ingestion, dbt for data models, and cloud-native orchestration. Set up cloud-agnostically, so the choice of GCP, AWS, Azure, or any other platform never becomes a constraint.

A data platform was only the beginning. The question many organizations will recognize: how do you add AI to this without building a second architecture?

Winparts' answer: AI jobs as a natural extension of the existing platform. Lightweight containers that read data from the warehouse, execute an LLM task, and write the result straight back. No separate tooling, no new infrastructure. The framework is set up once; after that you add use cases one by one.

The first use case: the returns process. Annotation at scale, pattern recognition, structural improvement points for assortiment and fitment data.

What you'll take away A hands-on session for anyone who wants to implement AI in an existing data platform — or hasn't started yet.

How to set up a data platform that is ready for AI, without starting over
How AI jobs work as a lightweight extension on an existing platform, cloud-agnostically
How to start with one concrete use case and scale from there
Honest trade-offs: which open source tools, which trade-offs, what it has concretely delivered