ETL, pipelines, architecture concepts
Master the Medallion Architecture with the water purification analogy. Bronze (raw), Silver (cleaned), Gold (business-ready) layers explained with format selection, transformation patterns, ownership model, cost optimization, and exact mapping to every pipeline we built on the blog.
Master every PySpark window function with real business scenarios. ROW_NUMBER, RANK, DENSE_RANK, LAG, LEAD, NTILE, running totals, moving averages, percent of group. Plus 5 real-world patterns: deduplication, gap detection, sessionization, YoY comparison, and top N per group.
Implement both SCD Type 1 and Type 2 using PySpark and Delta Lake MERGE in Databricks. Type 1: one MERGE with whenMatchedUpdate + whenNotMatchedInsertAll. Type 2: MERGE to expire + APPEND for new versions. Hash-based change detection, version numbering, idempotent tests, reusable functions, and complete Synapse Data Flow to PySpark MERGE mapping.
SCD Type 1 and Type 2 Using PySpark and Delta Lake MERGE in Azure Databricks Read More »
Master lazy evaluation — the most important PySpark concept nobody explains properly. Why Spark waits, how the Catalyst Optimizer rewrites your code, transformations vs actions with complete lists, predicate pushdown, column pruning, the DAG, narrow vs wide transformations, the cache trap, proving laziness with a hands-on experiment, and how it powers your SCD Type 2 pipeline.
The complete optimization playbook for Databricks. Small file problem and OPTIMIZE compaction, Z-ORDER for file skipping, conditional OPTIMIZE, VACUUM with retention and time travel interaction, partitioning strategy, AQE, broadcast joins, join best practices, caching, coalesce vs repartition, and the production checklist that turns 45-minute pipelines into 5-minute pipelines.
Master all 8 PySpark join types using the same two DataFrames. Inner, left, right, full outer, left_semi, left_anti, cross, and self joins with output tables, real-life analogies, SQL equivalents, broadcast optimization, duplicate column handling, and a decision table for choosing the right join.
Master Azure RBAC with the hotel key card analogy. Every role organized by service: Storage (the confusing ones explained), SQL, Synapse, Databricks, Data Factory, Key Vault, Networking, and Compute. Management plane vs data plane, four identity types, five real-world scenarios, the decision framework, principle of least privilege, and common 403 error fixes.
Master managed vs external tables in Databricks. Complete setup: Access Connector, Storage Credential, External Location, and external table creation. Proves data persistence after DROP TABLE with step-by-step walkthrough. Covers Delta operations on external tables, partitioning, VACUUM, granting access, and the three-layer Unity Catalog security model.
Build a production-grade unified pipeline that handles both full load and incremental load from a single config table. If Condition routes each table to the right strategy. Watermark pattern for incremental, date-partitioned outputs, complete audit logging, and zero-code-change table additions. The exact pipeline pattern used in enterprise data platforms.
Hands-on Delta Lake deep dive in Databricks. Every operation step by step: INSERT, UPDATE, DELETE, MERGE creating versions. Time travel three methods. Compare versions, track entities across history. RESTORE, VACUUM, Schema evolution, DeltaTable Python API.
