Overview

The Beltic Graph is a global relationship and intelligence engine that connects the world’s identity and risk data into one unified graph.
It allows financial institutions, fintechs, and regulatory platforms to visualize and understand the complex relationships between individuals, businesses, devices, and digital identifiers — in real time.

Built on top of graph database technologies, this system powers deep insights, risk linkages, and entity connections across the entire Beltic ecosystem.

Core Purpose

Beltic Graph enables the detection of hidden or non-obvious relationships that traditional databases cannot capture.
It aggregates and connects signals from a wide variety of sources, including:

• People — individual onboarding data, documents, behavioral metadata

• Businesses — corporate registration data, directors, shareholders, and partners

• Documents — identification and verification artifacts

• Contact Information — phone numbers, email addresses, and communication identifiers

• Digital Fingerprints — device, IP, and network data

• Regulatory Data — sanction lists, watchlists, AML and PEP data sources

• Legal Records — publicly available or customer-provided legal filings and compliance documents

By consolidating these data points, Beltic Graph allows teams to connect dots across entities, detect fraud faster, and understand global risk at a glance.

Architecture and Design

Unified Graph Model

The Graph product is built on top of a high-performance graph database layer, designed to handle unstructured, semi-structured, and structured data coming from over 150+ countries.
Each node and edge in the graph represents a meaningful real-world connection:

• A person node can connect to documents, addresses, or phone numbers.

• A business node links to shareholders, registration records, and counterparties.

• Cross-entity edges represent transactional, ownership, or behavioral relationships.

All entities are normalized into a single global schema that ensures data consistency, regardless of country or document origin.

Handling Global Unstructured Data

One of the most complex challenges Beltic Graph solves is harmonizing unstructured global data.

• Our ingestion layer applies NLP-based entity resolution, language normalization, and schema mapping to conform local formats (e.g., national ID formats, regional business registries) into a single unified model.

• The system continuously learns from new data sources, expanding entity coverage and improving deduplication accuracy.

• Multilingual and transliteration layers ensure that names, entities, and metadata align across languages and scripts.

Data Relationships and Inference

Beltic Graph’s intelligence layer infers relationships beyond direct matches:

• Shared identifiers: same phone number or email used across multiple applications.

• Common documents: repeated submission of the same ID or proof of address across different entities.

• Cross-ownership patterns: linking shareholders, directors, or beneficial owners across companies.

• Network overlap: matching devices, IP addresses, or behavioral patterns.

These inferred connections generate relationship scores and fraud likelihood indicators, helping analysts prioritize investigations or block high-risk activity automatically.

Example Use Cases

1. Fraud Detection

A bank notices multiple onboarding attempts using different names but the same phone number and address.
Beltic Graph highlights the linkage between these identities and flags them as part of a potential fraud cluster.

2. Sanctions & Watchlist Monitoring

A compliance analyst can query an entity (e.g., business director) and automatically view connections to known sanctioned entities or high-risk jurisdictions.

3. Document Reuse Tracking

When the same ID or proof-of-address document is uploaded by multiple applicants, the graph highlights the reuse event and its historical context — reducing false positives and enabling instant risk escalation.

4. Network Intelligence

By combining device fingerprinting and IP metadata, the system detects shared infrastructure patterns (e.g., multiple fraudulent businesses using the same hosting network).