Brickeer

Brickeer · Data and AI in production

Data and AI products people actually use.

We build data platforms, AI agents and models alongside your team. And we don't leave until they can run them without calling us.

Client work

Two products real people already use.

We built them with Containeer and Northa, and they are still running. You won't find their marketing numbers here, just the product as people use it.

Working in the public sector? Same method, with one more requirement: every decision the system makes has to be explainable.

See all client work →

How we work

Delivering is easy. What counts is that it's still in use six months later.

So from day one we sit with the people who will use it and show them every step before moving on.

What you'll notice day to day

Whoever leads your project looks at three things at once

Meet the team and how it works →
  • Technology

    Builds the system and leaves it running in your cloud.

  • Business

    Turns your goal into priorities and a way to measure whether it worked.

  • Market

    Thinks about who will use it and how it helps you sell more or spend less.

What we build

Systems your team can run.

Three practices that support each other. The diagrams you'll see are illustrations.

We work with the tools you already have
  • Databricks
  • BigQuery
  • Snowflake
  • S3 / R2
  • Airflow
  • Dagster
  • ADF
  • LangChain
  • LangGraph
  • RAG
  • Vector DB
  • MCP
  • Jupyter
  • MLflow
  • dbt
  • Apache Iceberg
See how it integrates with your stack →
01 · AI & AGENTSA person in charge

Agents and chatbots

One concrete task, its tools and a person where the answer matters: classifying, searching documents or leaving a draft ready to review.

What gets built
  • An agent on LangChain or LangGraph, with bounded tools and a defined input and output
  • Testing with your own examples, not a generic demo
  • Limits: what the system must not answer, and where it stops
LangChainLangGraphRAGVector DBMCP
See the practice →
A brick workshop: a central block linked to smaller ones holding a lens and a wrench, ending in a cream block where the result is reviewed.
OutcomeAn application the team uses in its work, with a way to know if it gets worse.
02 · ORCHESTRATIONAirflow · Dagster · ADF

Pipelines that can be run again

Airflow, Dagster, or ADF hold the schedule, the retries, and the alert when something fails. The pipeline can be started again without opening the laptop of the person who wrote it.

A sample pipeline

✓ All good
  1. IngestionDone

    Brings the data in from its source.

  2. ValidationDone

    Checks that the data has the expected shape.

  3. CleaningDone

    Removes duplicates and errors.

  4. ModelingDone

    Turns it into tables that can be queried.

  5. PublishingDone

    Leaves the result ready for your team to use.

Each step can be run again on its own.

Automated quality & testingSee the practice →
03 · LAKEHOUSEDatabricks · BigQuery · Snowflake · S3 / R2

Where the data already lives

Databricks, BigQuery or Snowflake, with the data on S3 or R2, depending on the account you already use. Spark provides the engine when you need to process at scale. The data stays where it already is.

Your platform, in your cloud

The data lives in your own buckets. It is yours and it can move.

AWS S3 · Google Cloud Storage · Cloudflare R2

Open formats from start to finish, without tying you to a vendor.

Compute in your own cloudSee the practice →

AI + people

AI speeds things up. A person decides.

We use AI for the repetitive parts: classifying, searching, writing the first draft. Whatever gets approved is approved by a person.

What AI speeds up

  • Classifying, searching and drafting over your own documents
  • Repeating tasks that a pipeline can run again
  • Spotting the odd case so someone looks at it

What a person decides

  • The scope, the priorities and the changes
  • What gets approved before touching production data
  • What to do with a recommendation, and what not to take as certain

Use case

Building a data architecture, brick by brick

AI proposes

Builds the dashboards and the alerts.

A person decides

Decides what is measured and what to do about it.

Illustrative case. AI proposes; a person decides.

An illustrative, summarized case. Decisions are always made by a person.

Start to finish

Five phases. You always know where the project stands.

Initiation, plan, execution, control and close. In «Method» you see who does what in each phase and what you get.

01

Initiation

A sponsor as first contact, the decision that has to improve and a success criterion. If any of them is not clear yet, we define it together before starting.

See how we work →

Let's talk

Shall we start building?

Tell us what you want to achieve. If it is not all defined yet, we will work it out together.

No formal brief needed. Tell us what you want to achieve and what you have today, and we'll work out together how to shape it.

Direct channel

The form opens your mail program. There is no inbox on this site: you send the message yourself to info@brickeer.com.

It's read by the people who would build it.

How we work with you
  • A direct reply in under 24 business hours.
  • We tell you honestly whether we can help.
  • You talk to the person who will build it.

Step 1 of 4

What do you need?

Pick the closest one. If you are not sure, pick «Other».

What do you need?