AI-Driven Cyber Crisis Simulation (PoC)
A strategic "serious game" where a CTO defends a FinTech infrastructure against AI-driven ransomware. Players must balance real-time security trade-offs (isolation, patching, scanning) against business constraints like revenue flow, GDPR compliance, and reputation.
Multi-Agent (LangGraph): Orchestrates specialized CISO, SRE, and Hacker nodes to manage complex, cyclical reasoning.
Expert RAG (FAISS): Grounded in real-world knowledge—MITRE ATT&CK tactics, SRE patterns, and European regulatory frameworks.
Semantic Caching: Purifies game states into canonical forms to identify identical strategic situations, slashing inference costs.
Zero Hallucination: A deterministic engine using Pydantic for 1:1 UI mapping and structured data integrity.
Language: Python
AI Frameworks: LangChain, LangGraph
Vector DB: FAISS
Models: OpenAI API, Pydantic (State Validation)
Visualization: Mermaid.js
Deterministic Logic: Outcomes derive strictly from infrastructure traits and revenue flows—no random chance (RNG).
16 Specialized Actions: Fully implemented across security and audit roles.
Performance: Semantic cache achieves >0.99 similarity for strategic state purification.