Founder & Agentic Systems Engineer at Shadow Auditor AI
Building Autonomous Multi-Agent Swarms to Eliminate LLM Hallucinations in Offensive Cybersecurity.
I am a 16-year-old cybersecurity researcher, AI engineer, and agentic systems architect based in Brussels, Belgium. I specialize in designing deterministic AI workflows for complex code analysis and automated vulnerability detection.
Instead of building generic wrapper applications, I focus on building resilient orchestration frameworks, abstract syntax tree (AST) parsing, and secure sandbox containment to engineer production-ready security tools.
"Oh, so you are a vibe coder?"
Yes, I am. But there is a fundamental difference. I don't just prompt and hope for the best; I architect multi-agent systems, enforce strict anti-hallucination protocols, and design rigorous quality and security gates into software development.
"Why don't you learn traditional coding?"
I did, but with a different philosophy. I find it highly inefficient to waste time memorizing pure syntax in 2026 and beyond. If your vision of the future is still focused on writing code by hand, you will be replaced.
I chose to master the structural logic behind software. I design, audit, and make the high-level architectural decisions, while AI handles the execution loop. Being born into the AI era requires realism: spending years learning language-specific syntax is obsolete for my generation. We build systems, not just lines of code.
| Domain | Technologies & Frameworks |
|---|---|
| AI & Orchestration | LangGraph, LangChain, Multi-Agent Architecture, Vector Embeddings (Ollama/OpenAI) |
| Cybersecurity / SAST | AST Parsing (Tree-sitter native loading), Taint Tracking, CWE Classification, DAST Containment |
| Backend & Systems | Node.js (v24+), TypeScript, Docker Isolation, Bash/PowerShell scripting |
| CI/CD & DevOps | GitHub Actions, Automated Release Gates (Ubuntu & Windows runners), SARIF Compliance |
My primary development effort is Shadow Auditor, an autonomous AI-powered security analysis engine.
- Multi-Agent Collaboration: Employs a 7-role swarm architecture over a shared Blackboard state (Recon, TaintTracer, ExploitAnalyst, Verifier, Reporter, Orchestrator, Patch Engineer).
- Anti-Hallucination Choke Point: Integrates a hostile Devil's Advocate agent node that actively challenges and filters out false positives before reporting.
- Enterprise Ready: Full static validation, deterministic pipeline handoffs, and content-derived stable SARIF rule indexing.
- LinkedIn: Yahya Essabbahy
- Startup Email: contact@shadow-auditor.app
“Autonomous security for teams that demand evidence.”