Skip to main content

AI-AML-SENTINEL: Product Requirements Document (PRD) v1.0

1. Executive Summary​

AI-AML-SENTINEL is an intelligent compliance engine designed to augment traditional transaction monitoring systems. It uses Large Language Models (LLMs) to identify complex, non-linear financial crime patterns and provides actionable compliance advisory in real-time.

2. Problem Statement​

Traditional AML systems rely on hardcoded rules (e.g., "if amount > $10,000, then flag"). These systems suffer from:

  • High false positive rates.
  • Inability to detect "grey area" risks or evolving money laundering typologies.
  • Lag in adapting to new regulatory standards (SFC, MiCA, MAS).

3. Core Features​

3.1 Semantic Risk Analysis (AI-Core)​

  • Typology Mapping: Automate the identification of patterns like "Layering" and "Integration" by analyzing transaction narratives and metadata.
  • Contextual Scoring: Assign risk scores based on the "intent" and "structure" of transactions, not just thresholds.

3.2 Regulatory Intelligence Integration​

  • Dynamic Benchmarking: Automatically update internal risk parameters against the latest SFC VATP guidelines and MiCA CASP requirements.
  • Audit-Ready Justification: Generate human-readable explanations for every flag to assist compliance officers during regulatory audits.

3.3 Multi-License Support​

  • Unified monitoring for entities holding both crypto (VASP) and fiat (Payment) licenses.

4. Technical Architecture (High-Level)​

  • API Layer: FastAPI for high-speed transaction ingestion.
  • Intelligence Layer: Google Gemini 1.5 Pro / Flash for semantic analysis.
  • Data Layer: Vector database (e.g., Pinecone/Chroma) for storing regulatory knowledge and historical patterns.

5. Target Users​

  • Licensed VATPs in Hong Kong.
  • CASPs in the European Union.
  • Cross-border payment processors.

Status: Draft - 2026-04-24 Owner: Louie / UWAY Innovation Limited