Syenah Risk Sentinel fuses Generative AI, GraphDB-based Knowledge Graphs, Supply Chain intelligence, Graph Analytics, and evidence-based GraphRAG in a single platform.
Today’s risk teams use multiple disconnected tools. Data doesn’t flow between systems.
Analysts drown in alerts, manually connecting dots across siloed systems. Critical risks propagate faster than spreadsheets can track. You pay €150K–300K+/year for intelligence that still can’t see the connections.
Six technology layers work together.
Generative AI detects business and sustainability incidents from unstructured sources. These incidents feed a knowledge graph where entities and relationships are resolved. The graph enables supply chain traversal and network analytics. GraphRAG turns natural language questions into evidence-based answers. Together, the layers deliver complete risk intelligence in one platform.
We don’t buy API calls from OpenAI.
We run open-source LLMs on our own dedicated NVIDIA DGX hardware. Complete control over model behaviour, zero data leakage, and 90% lower inference costs.
Multi-stage NLP pipeline processing 1M+ articles daily.
Open source models 10+ specialised tasks per article, from incident detection to severity scoring, outputting structured JSON directly to our knowledge graph.
| Component | Technology | Specification |
|---|---|---|
| Primary Model | Llama-4-Maverick-17B-128E | Complex tasks requiring reasoning |
| Classification Model | Llama-4-Maverick-17B-128E | Multi-label classification, 256 risk factors |
| NER/NEL | Llama-4-Maverick-17B-128E | 2.2M+ entity knowledge base |
| Embedding Model | Multilingual-e5-large | Semantic search, similarity matching |
| Severity Extraction Model | Llama-4-Maverick-17B-128E | Complex tasks requiring reasoning |
Native analysis in 100+ languages, not translated then analyzed. Preserves legal terminology and cultural context.
Lid.176.binHigh-precision classification to identify materially relevant ESG and business incidents, distinguishing real-world events (e.g., factory fires, labor protests) from non-incidents such as drills, announcements, or routine communications.
The engine analyzes the core narrative to extract structured data, information and context, transforming unstructured text into database-ready fields.
Maps ambiguous actor names (e.g., “Amazon” vs. “Amazon River”) to a verified global entity DB of 2.2M companies. This process resolves subsidiaries, aliases, and operating units to their ultimate parent organisation.
Content is evaluated against our proprietary 256-Factor Taxonomy, assessing each incident for specific risks across four strategic pillars. Every risk factor is mapped to global regulatory frameworks to enable automated compliance reporting.
Context-aware, multi-dimensional scoring that evaluates the true impact of each incident across key risk dimensions, distinguishing minor issues from critical crises with consistent, auditable rigor.
TigerGraph-powered Knowledge Graph storing 2.2M+ entities, 1B+ relationships, and real-time incident data.
Multi-hop graph traversal in under 100ms, map every connection to sanctioned entities, high-risk suppliers, or flagged counterparties across your entire exposure. Queries that SQL can’t express, answered before you finish asking.
| Node Type | Count | Key Properties |
|---|---|---|
| :Company | 2.2M | id, name, aliases[], industry, country, risk_score |
| :Person | 2.1M | id, name, roles[], pep_status, sanction_status |
| :Incident | 4.2M | id, summary, severity, risk_factors[], date, sources[] |
| :Location | 890K | id, name, type, country, coordinates |
| :Article | 45M | id, url, title, published_at, source, language |
Optimized for real-time risk analysis and deep supply chain exploration across large enterprise graphs.
Retrieve profiles, risk indicators, sanctions, and key relationships in real time
Explore upstream and downstream exposure with controllable depth, even for the largest global enterprises (<4s)
Trace how incidents and risk events spread through ownership and supply chain networks
Detect direct and indirect connections to sanctioned or restricted entities
Identify and explain why two entities are connected, including intermediate nodes and relationships
Continuous ingestion across structured data, unstructured text, and AI-generated signals.
Incidents, sanctions, and relationship updates become queryable with low latency
AI-extracted entities, relationships, and external risk signals streamed into the graph within seconds
Continuous merges without full reloads or analytical resets
Every relationship and risk signal is linked to its source and supporting evidence
Updates immediately usable by AI agents and graph-based RAG workflows
Built for large-scale, always-on supply chain and risk intelligence workloads. A production-grade graph infrastructure designed to support real-time traversal, high-concurrency access, and continuous data growth without sacrificing performance or availability.
Multi-node enterprise graph cluster with fault tolerance and automatic failover, ensuring uninterrupted access to critical risk insights.
Designed for concurrent analytical and exploratory queries across large, highly connected graphs.
Large in-memory footprint per node with NVMe-backed storage for fast traversal and predictable latency.
Support for complex traversals, path analysis, and network analytics required by supply-chain and risk use cases.
A high-fidelity relationship model connecting companies, incidents, and risk signals across ownership, Supply chain, financial exposure, and real-world events.
Direct links capturing ownership exposure, operational involvement, and downstream impact from ESG, compliance, and risk events.
Multi-layer supplier and customer connections, with optional access to transaction-level shipment details (time, origin, destination, value, goods).
Subsidiaries, affiliates, investment arms, and merged entities, preserving both control and economic relationships.
Investments, creditors, borrowers, strategic alliances, sponsors, and investor-related service providers.
Franchise, licensing, leasing, landlord-tenant, and other long-term contractual relationships.
Direct connections between companies and sanction lists or regulatory sources.
Algorithmically inferred relationships capturing structural proximity and ecosystem membership.
Incidents connected to countries, industries, risk factors, breaches, NGOs, and source documents.
Text units, extracted entities, extracted relationships, and semantic communities enabling graph-based RAG and explainable retrieval.
Graph Data Science algorithms for real insights: measure centrality, detect communities, trace paths, and find anomalies. Explore everything with an interactive graph visualization tool.
Identify critical nodes and influence patterns in the network.
Discover clusters of related entities for group risk assessment.
Trace connections between entities through any relationship type.
Identify unusual patterns that may indicate hidden risks.
Web-based graph explorer for investigative analysis. Point-and-click traversal, filtering, and export.
Hybrid retrieval combining vector similarity search with graph traversal. Natural language queries return grounded answers with citations, zero hallucination by design.
| Stage | Technology | Output |
|---|---|---|
| 1. Query Understanding | LLM intent classification + NER + NEL | Intent, entities, constraints |
| 2. Query Embedding | Text-embedding-3-small | Query vector |
| 3. Retrieval | TigerGraph | Top-100 candidate documents |
| 4. Graph Expansion | GSQL | Related incidents, relationships |
| 5. Response Generation | Llama-4-Maverick-17B-128E | Grounded answer + sources |
Combines semantic similarity with graph structure for contextually relevant results.
Every factual claim linked to source evidence. Full audit trail for compliance.
Retrieval-constrained generation ensures factual grounding.
Flexible response types from conversational answers to structured reports.
End-to-end flow from raw article to queryable intelligence in under 5 minutes
No need to take our word for it. Explore the platform at your own pace and see the value for yourself.
No commitment. Full access. Enterprise-ready.
