Three courses, depending on where you start
You learn by building something that has to work. Pick your starting point and join the waitlist: people on the list hear about dates first.
Price and dates: to be confirmed. People on the waitlist get priority and launch pricing.
One path, three steps
From building reports to modeling the data
01 · Analytics Engineering course →From coding APIs to building pipelines
02 · Modern Data Engineer course →From moving data to preparing data for AI
03 · AI-native Data Engineer course →
Analytics Engineering course
You go from building reports to building the data model behind them, tested and documented.
What you work with
- SQL
DuckDB
dbt
Git
- BI
Who it is for
- You work with Excel, SQL or Power BI
- You are a data analyst or you build the reports
- You want to go from asking for data to building the model the whole team uses
What you need to know
- Excel or basic SQL
- No programming needed
See syllabus · 5 weeks
- Analytical SQLWindow functions, CTEs and aggregations with DuckDB. Why a transactional database is not built for analysis.
- Data that arrives rightIncremental, idempotent loads. The interview question: what happens if it runs twice?
- Dimensional modelingFacts, dimensions and star schemas. How to keep the history of a customer that changes.
- Transformation with dbtStaging, marts, refs, tests and docs. Git and code review, like a real data team.
- Metrics and BIOne definition per metric and a dashboard connected to the model, not to a spreadsheet.
What you take away
A tested dbt project on a fictional digital wallet case, 10 business questions answered in SQL and a dashboard that reads from the model.
Modern Data Engineer course
You go from thinking in requests to thinking in data, and leave with an end-to-end data platform you can defend in an interview.
What you work with
Python
DuckDB
Postgres
Parquet
dbt
Iceberg
Airflow
Who it is for
- You are a mid or senior backend or full-stack developer
- You want to move into data without starting from scratch
- At your company or in a remote role
What you need to know
- Python in production (or Java, Node or Go and you are fine with Python)
- Basic SQL
- Git and Docker as a user
See syllabus · 5 weeks
- Analytical SQL and the data mindsetOLTP vs OLAP, window functions and CTEs with DuckDB and Postgres.
- Incremental, idempotent ingestionPulling from an API with cursors and retries, and loads you can rerun without duplicates. Python, Parquet and partitions.
- Modeling with dbtStar schema, facts and dimensions, SCD type 2. Staging, marts and docs.
- Lakehouse vs warehouseParquet and Iceberg on object storage, when BigQuery or Snowflake makes sense and what each path costs.
- Orchestration and qualityAirflow with dependencies, backfills and alerts. dbt tests and data contracts: find out before the business does.
What you take away
A public repo with ingestion, an Iceberg lakehouse, a dbt model, Airflow orchestration and tests, plus an ADR explaining every decision.
AI-native Data Engineer course
Your RAG does not fail because of the model, it fails because of the data. You build the data layer for AI: from documents to an index that refreshes itself and can be measured.
What you work with
Kubernetes
MinIO
pgvector
DuckDB
Debezium
Spark
Grafana
Who it is for
- You are a data engineer with 2+ years and pipelines in production, or a data tech lead
- You are being asked for "something with AI": RAG, vector DBs, unstructured data
What you need to know
- Advanced SQL and modeling
- Fluent Python, Docker and Git
- Experience orchestrating pipelines in production
- No Kubernetes needed: we start from zero
See syllabus · 6 weeks
- A data platform on KubernetesPods, jobs, volumes and Helm. A local cluster with MinIO, Postgres with pgvector and DuckDB.
- Unstructured ingestionParsing PDFs, HTML and tables, metadata, lineage and deduplication. What gets lost when parsing.
- Chunking and embeddings as modelingChunking strategies, open source vs API models, pgvector vs Qdrant and hybrid search.
- Streaming for a fresh indexCDC with Debezium, Redpanda and Spark Structured Streaming. Re-embed only what changed.
- Retrieval evaluationGolden set, recall@k and RAGAS. Evaluation runs in CI and blocks the change if quality drops.
- Cost, observability and productionCost per 1,000 documents, metrics with Prometheus and Grafana, and an agent that uses the index as a tool.
What you take away
A documents-to-vector-index pipeline that runs on Kubernetes, updates in streaming and tracks quality and cost per document. With your own numbers to defend each decision.
How the format works
- Live, cohort-based classes: concept and demo, plus a guided lab.
- 5 to 6 hours per week, designed for people working full time.
- One case through the whole course: each week adds a layer to the same repo.
- Every class is recorded.
- Limited seats, so each person's work gets reviewed.
- In Spanish, with banking and fintech cases.
Price and dates: to be confirmed. People on the waitlist get priority and launch pricing.