RiskTerrain
ai

RiskTerrain

S&P 500 Risk Intelligence

RiskTerrain is an AI-powered risk intelligence platform that maps how geopolitical disruptions cascade through S&P 500 supply chains. A 6-node LangGraph pipeline ingests real-time events from USGS earthquake feeds and NewsAPI, traverses a supply chain knowledge graph of 154 companies stored in SurrealDB, and scores cascading risk exposure — all in under 8 seconds. The frontend renders an interactive 3D risk terrain using Three.js.

Key Features

6-node LangGraph pipeline for event ingestion, classification, and risk scoring
154 S&P 500 companies modeled as a supply chain knowledge graph in SurrealDB
Multi-hop graph traversal to trace cascading risk through supplier networks
Hybrid graph-vector retrieval combining structural queries with semantic search
sentence-transformer embeddings for event-to-company relevance matching
Sub-8-second end-to-end latency from event ingestion to risk score output
3D risk terrain visualization built with Three.js

LangGraph Pipeline

The pipeline consists of 6 sequential nodes: event ingestion (USGS + NewsAPI), event classification (type, severity, region), entity extraction (affected companies/regions), graph traversal (multi-hop supply chain propagation), risk scoring (weighted by proximity, dependency strength, and event severity), and output formatting. Each node is independently testable and the pipeline supports both batch and streaming modes.

Supply Chain Graph

SurrealDB stores a multi-relational graph of 154 S&P 500 companies with edges representing supplier, customer, and competitor relationships. The graph includes geographic metadata enabling region-based traversal — when an earthquake hits Taiwan, the system traces semiconductor supply chains to identify exposed companies across multiple hops.

Hybrid Retrieval

The system combines structural graph queries ("find all companies within 2 hops of affected suppliers") with semantic vector search ("find companies whose business descriptions are relevant to this event type"). This hybrid approach catches both direct supply chain exposure and indirect thematic risk that pure graph traversal would miss.

Tech Stack

Python
LangGraph
SurrealDB
Three.js
Docker