Fraud labels arrive late and online features disagree with training data.
You own the technical recommendation and the evidence behind it. A reviewer must be able to reproduce the decision, identify which assumptions are estimates, and operate the failure path without asking the author.
Versioned events or rows with entity ID, event time, availability time, payload, and mature outcome where applicable.
A registered artifact plus evaluation report, decision policy, architecture trace, observability plan, and rollback procedure.
Beat the declared baseline without violating latency, cost, capacity, subgroup, or freshness guardrails.
Submit evidence another engineer can inspect.
- 01
Contracts and lineage
Include assumptions, identity/version fields, one success example, one counterexample, and the command or query used to verify it.
- 02
Temporal join specification
Include assumptions, identity/version fields, one success example, one counterexample, and the command or query used to verify it.
- 03
Late-event/backfill policy
Include assumptions, identity/version fields, one success example, one counterexample, and the command or query used to verify it.
- 04
Offline-online parity report
Include assumptions, identity/version fields, one success example, one counterexample, and the command or query used to verify it.
The result must stay true after packaging.
- All training inputs are knowable at the historical decision cutoff.
- Raw-input inference after serialization matches the validated reference within declared tolerance.
- Every artifact and decision can be joined to code, data, features, configuration, and model version.
- A failed retry is idempotent and cannot publish partial output as complete.
- Quality is reported by temporal, operational, and affected-user slices—not aggregate only.
- The rollback or fallback has a tested trigger, owner, and bounded recovery time.
Try to falsify your own recommendation.
- Replay one hand-calculated input and one production-shaped batch.
- Inject a missing, late, malformed, and out-of-distribution input.
- Restart at the worst possible commit boundary.
- Load-test the constrained path, including cold start and a failed dependency.
- Compare one protected or high-risk slice and document uncertainty.
- Have the fallback produce an auditable result before calling the system ready.