MERGE / Upsert Correctness Auditor
Data Engineering
Prove a warehouse MERGE or upsert cannot duplicate keys, drop deletes, or race with concurrent writers under real source mutations.
mergeupsert+2
00
Loading prompts…
Curated AI prompts for data engineering, software engineering, and AI engineering.
Data Engineering
Prove a warehouse MERGE or upsert cannot duplicate keys, drop deletes, or race with concurrent writers under real source mutations.
Data Engineering
Find overlapping intervals, missed closes, and point-in-time join bugs in a slowly changing dimension under corrections and late source rows.
Data Engineering
Design watermarks, allowed lateness, and side outputs so windowed streaming jobs stay correct when events arrive late, out of order, or after idle gaps.
Data Engineering
Design a restartable SaaS API extractor that survives pagination, rate limits, partial pages, and vendor cursor resets without duplicating or skipping records.
Data Engineering
Review Iceberg or Delta snapshot, isolation, and planning behavior so concurrent writers, time travel, and compaction cannot serve torn or expired data.
Data Engineering
Audit a metrics or semantic layer so revenue, conversion, and other executive numbers have one grain, one join path, and no silent double-counting.
Data Engineering
Attribute warehouse, lake, and streaming spend to teams and data products with guardrails that cut cost without silently dropping SLAs.
Data Engineering
Design a deterministic-then-probabilistic identity pipeline with reversible merges, audit trails, and replay when source keys collide or split.
Data Engineering
Build CI and runtime gates so Avro, Protobuf, or JSON Schema changes cannot break consumers, sinks, or warehouse loaders after they ship.
Data Engineering
Design a durable S3/GCS/ADLS landing zone that handles late files, partial uploads, duplicate drops, and schema-mixed folders without poisoning downstream tables.