Know the tool.
Understand the work.
Ten connected areas of practice. Choose a question worth answering, then develop the technical evidence to answer it.
Find your starting point →
Choose the question.
Then choose the tools.
A skill area is a place to investigate, not a checklist of products to collect. Each section connects technical ideas to an assignment and available practice.
Start with the difficulty that fits your prerequisites. An experienced visitor can go directly to advanced project work without repeating introductory reading.
Work through an assignment →Refresh concepts, follow an example, reconcile your first result.
Combine tools, investigate exceptions and produce a reproducible assignment.
Challenge assumptions, evaluate failure modes and defend design decisions. Start here when you have the prerequisites.
SQL & data engineering
SQL, CTEs, window functions, query plans, data modeling, ETL/ELT and pipeline quality.
Professional application: Reconcile order-grain totals before joining item-level data. Compare a window calculation against an independently grouped control.
FROM / JOINEstablish source grain and join cardinalityWHEREKeep eligible input rowsGROUP BYAggregate at the intended reporting grainHAVINGFilter grouped resultsSELECTChoose output expressionsORDER BYMake presentation order explicit
A one-to-many join can inflate a sum before aggregation. Check key uniqueness and reconcile totals before interpreting the output.
Python for data analytics
Pandas and NumPy, cleaning, APIs, automation, statistics, reproducible notebooks and tests.
Professional application: Validate identifiers and money types, preserve rejected rows and automate a report with a source fingerprint.
Business intelligence
Power BI, Tableau, semantic models, DAX, KPI definitions and accessible dashboard design.
Professional application: Define each KPI at its source grain. Reconcile the model before choosing a visual and document filter behavior.
Machine learning
Regression, classification, clustering, forecasting, feature engineering, leakage, uncertainty and monitoring.
Professional application: Write a prediction-time contract, split before preprocessing and compare a simple baseline on held-out observations.
The linked companions support reproducibility and interpretation; the course catalogue shows available ML modules and access requirements.
Generative AI & AI engineering
LLMs, prompts, retrieval, embeddings, vector search, agents, tool contracts and evaluation.
Professional application: Build a bounded document-retrieval experiment. Test answerable and unsupported questions and require human review for consequential actions.
Prepare permitted documents
- Parse and split documents
- Embed chunks
- Store vectors with source IDs and access metadata
Receive a question
- Apply access and input controls
- Embed the query with a compatible model
- Retrieve authorized relevant chunks from the index
Document index + query embedding → authorized retrieval → selected evidence
- Assemble evidenceProvide retrieved context and source identifiers to the model
- Generate or abstainReturn a bounded answer or say the evidence is insufficient
- Evaluate and reviewCheck retrieval, citation support and failure cases; use human oversight
Retrieval does not guarantee relevance, and citations do not prove an answer is correct. The linked companion starts with local retrieval evaluation; no paid model call is required.
Cloud data platforms
Storage, warehouses, orchestration, containers, deployment, MLOps, monitoring, access and cost controls.
Professional application: Map service responsibilities, test a local release and rollback, and document permissions before choosing a cloud provider.
Business analytics & decisions
Requirements, root-cause hypotheses, statistical reasoning, experiments, stakeholder communication and decision memos.
Professional application: Separate an observed change from a causal explanation. Specify the decision, alternatives and evidence needed next.
Data governance & responsible AI
Privacy, access, lineage, quality controls, audit trails, bias reviews and responsible automation.
Professional application: Trace a reported number to its source and document exclusions. Identify sensitive fields and a human approval boundary.
Industry-specific analytics
Retail, healthcare, construction, finance, education, nonprofits, logistics, manufacturing and hospitality.
Professional application: Start with available fictional retail and inventory cases. Other sectors are exploration directions, not a claim of released labs or client work.
Professional practice
Technical reviews, portfolio projects, code review, business cases, interview preparation, Git and analytical communication.
Professional application: Produce a reproducible evidence folder and a concise decision note. Ask a reviewer to challenge a limitation, not simply admire the chart.
Understanding and application are different evidence.
A correct explanation is a start. A checked output, a reproducible workflow and an honest account of limitations show what you can apply.
Explore Professional Practice Labs →