Shape
Required fields, nested structures, field presence and representative type checks.
DATA QA / VALIDATION
Make bad data fail loudly and repeatably.
JSON, JSONL and CSV checks for exports, APIs and automation pipelines. My AI/LLM-assisted development practice has run continuously since November 2022 and includes data QA, validation tooling and CI-oriented handoff.
What gets checked.
Quality
Duplicates, malformed values, invalid rows and explicit treatment of unknowns.
Repeatability
Reusable validator or command plus human-readable results and deterministic exit behavior when appropriate.
Inspectable implementation.
Data Shape Guard
Dependency-free drift detection and strict JSON/JSONL validation with tests and CI.
Have a messy export or unstable payload?
Send a sanitized representative sample and the expected shape.