Project 1 · Data systems0/5 completeData Contract Lab
A document classifier cannot learn from a row with no stable identity, empty text, absent label, or evaluation reservation. This project turns 30 messy rows into a release later projects load by ID.
- L1Row Contract & Validatormake one row trustworthy · ~12 minstart →
- L2Deduplication + Provenancenormalize before you dedupe · ~15 minlocked
- L3Group-Safe Splitsplit by group, never by row · ~18 minlocked
- L4Incremental Ingest + Dead Letterscommit only what is durable · ~20 minlocked
- L5Rebuildable Dataset Releasecontent is the identity · ~25 minlocked
Artifacts this project producesL1 · Row Contract & Validatoroutputs: row-contract
L2 · Deduplication + Provenanceoutputs: dedup-report
L3 · Group-Safe Splitoutputs: split-policy
L4 · Incremental Ingest + Dead Lettersoutputs: checkpoint
L5 · Rebuildable Dataset Releaseoutputs: dataset-release