Loadingone moment…
D8Loop · Field notes
Read the concepts, patterns, and interview signals behind every module — 11 knowledge bases, read like an archive. Scroll to move through them; jump from the index on the left.
Grain, keys, ownership, and governance decisions that make data products durable instead of merely diagrammed.
Ingestion, orchestration, lakehouse choices, streaming edges, and failure paths for systems that must be operated.
Spark internals interviews actually probe: jobs, stages, and tasks; shuffles and skew; partitioning, caching, joins, and the memory model.
Relational thinking, analytical query patterns, and the habit of explaining results before optimizing syntax.
Transformation code that stays legible under messy inputs, edge cases, and interviewer follow-up questions.
Scheduling semantics that trip everyone up, DAGs that survive review, and the concurrency, executor, and failure-mode decisions that separate a scheduled script from an orchestrated system.
Models, refs, and materializations; tests, contracts, and lineage — the transformation layer most warehouses are actually built on, and the half of the stack Airflow triggers.
A drilled bank across eight subjects, organised by the form the question takes rather than the topic it covers — because writing a query, predicting its output, and finding out why it returns the wrong number fail separately.
End-to-end builds that double as portfolio proof: ingestion, transformation, orchestration, and serving on realistic datasets and stacks.
What the job assumes you already know and no chapter ever teaches: the git commands before your first pull request, the shell you work in, the words a team uses on day one. Looked up in a minute, not studied.
How the interview actually runs, at 50 named companies: every round, the questions candidates were asked, what got people rejected, and what to prepare first — plus six chapters on the rounds themselves.