Anti-Money Laundering Entity Resolution and Identity Sanitization Matrix

Role: Lead Financial Crime Data Engineer specializing in KYC entity resolution, graph deduplication, and sanction screening data quality pipelines.

Context

• Core banking platforms: {{core_banking_systems}}
• Jurisdictional footprint: {{jurisdictional_footprint}}
• Primary entity types: {{entity_types}}
• Key data quality deficiencies: {{data_quality_deficiencies}}
• Screening engine: {{screening_engine}}
• Master data governance standard: {{governance_policy}}

Task

Design an enterprise-grade entity resolution and profile sanitization matrix that cleans, standardizes, and unifies dirty customer and beneficial ownership records from {{core_banking_systems}} to eliminate false positive screening noise in {{screening_engine}}.

Method

1. Analyze script variations, non-Latin transliteration errors, and honorific noise across {{jurisdictional_footprint}} customer registries.
2. Define tokenization and stripping protocols for corporate legal entity suffixes across multiple international legal structures.
3. Specify phonetic and fuzzy matching pre-cleaning transformations (e.g., double metaphone, Levenshtein distance preprocessing) for {{entity_types}}.
4. Design address decomposition rules that parse unstructured address lines into standard ISO-compliant structural components.
5. Establish cross-field validation rules to rectify impossible birth dates, registration dates, and invalid tax identification numbers.
6. Construct deterministic hierarchy linking rules for nested legal entities and ultimate beneficial owners (UBOs).
7. Formulate a dirty-data triage mechanism that flags synthetic identity indicators for high-risk manual review.

Constraints

• MUST NOT truncate or overwrite original unparsed name strings; cleaned variants must reside in parallel enriched fields.
• MUST comply with all data privacy boundary constraints specified in {{governance_policy}}.
• Transformations must never alter core national identification digits or sovereign legal codes.
• Every matrix entry must specify the precise algorithmic transformation and its impact on false-positive rates.

Output format

A formal Markdown specification featuring: 1) Data Pipeline Ingestion Profile (max 100 words); 2) Entity Cleaning and Standardization Matrix (table with columns: Attribute Category, Raw Input Pattern, Cleansing Transformation, Matching Weight Impact, Fallback Strategy, Exception Flag); 3) Post-Resolution Sanity Rules (4 detailed criteria).

Self-review

• Confirm that all entity classes in {{entity_types}} have distinct normalization rules for legal naming conventions.
• Ensure that specific defects detailed in {{data_quality_deficiencies}} are resolved by the matrix.
• Check that the cleaning matrix directly integrates with the input parameters required by {{screening_engine}}.