Build a Multi-Source Watchlist Lineage and Adjudication Dispute Arbiter

Ingest divergent sanctions screening dispositions from multiple junior reviewers across distinct watchlist feeds. Map entity provenance to FollowTheMoney schema models and compile evidence arbitration packages for chief compliance officers.

Sanctions Review Escalation Lead

Ingest conflicting analyst screening adjudications containing individual reviewer rationale and confidence scores.

Retrieve upstream entity records and lineage citations from local FollowTheMoney entities via a simulated yente API.

Detect discrepancies in underlying sanction program tags, aliases, and source update timestamps across conflicting reviews.

Synthesize an arbitration dossier compiling provenance lineage and consensus recommendation drafts for senior auditor sign-off.

A multi-column discrepancy dashboard showing reviewer adjudications alongside underlying FollowTheMoney entity source provenance trees.

Normalized JSON review decision logs cross-referenced with synthetic FollowTheMoney entity payloads hosted on a local test mock server. Starter examples below are synthetic. They are not downloaded records.

OpenSanctions data toolkit Leverages FollowTheMoney schema entity attributes, sanctions program properties, and yente matchmaking endpoint structures.

dspy, pydantic, fastapi. DSPy compiles robust prompt pipelines that objectively synthesize conflicting analyst rationale into standardized dispute evidence briefs. Local or hosted LLM evaluated with teleprompters for zero-shot text summarization and factual consistency.

Submit runnable source, pinned dependencies, setup commands, a working interface, example outputs, and a recorded end-to-end demo. Include tests and an action log.

Extracts source provenance identifiers and dataset publication timestamps for each contested watchlist hit.

Identifies exact root causes of reviewer divergence including alias coverage or stale watchlist versions.

Outputs an immutable arbitration dossier with deterministic discrepancy scoring and zero unassigned review items.

```json [ { "input": { "entity_id": "ftm-person-8841", "reviews": [ { "analyst": "analyst_a", "disposition": "FALSE_POSITIVE", "notes": "Birth year differs by 10 years" }, { "analyst": "analyst_b", "disposition": "POTENTIAL_MATCH", "notes": "Matches alternate spelling and nationality" } ] }, "expected_output": { "dispute_status": "ESCALATED_FOR_ARBITRATION", "provenance_dataset": "ofac_sdn", "root_cause": "BIRTH_DATE_ALIAS_CONFLICT", "provenance_timestamp": "2024-03-01T00:00:00Z", "consensus_score": 0.5, "recommended_review_step": "REQUEST_PASSPORT_COPY" } }, { "input": { "entity_id": "ftm-corp-1029", "reviews": [ { "analyst": "analyst_a", "disposition": "TRUE_POSITIVE", "notes": "Exact registration number match" } ] }, "expected_output": { "dispute_status": "UNANIMOUS_CONSISTENT", "provenance_dataset": "eu_fsf", "root_cause": "NONE", "provenance_timestamp": "2024-02-15T00:00:00Z", "consensus_score": 1, "recommended_review_step": "APPLY_REGULATORY_RESTRICTION" } } ] ```

Performs advisory compliance hit adjudication without autonomous account closures, asset freezes, or live law enforcement reports. Use mock external writes. Require human review before any real deployment.