Quality Assurance Engineer → Cybersecurity & AI Red Teaming (LLM Security)
📍 Eden Prairie, MN
📧 rodney_stanley@hotmail.com
🔗 LinkedIn

I bring 20+ years of experience delivering high-quality, secure software and now apply that testing discipline to cybersecurity and AI red-teaming.

My focus is on systematically identifying failure modes in AI systems, documenting risk, implementing mitigations, and validating defenses through structured adversarial testing. I am especially interested in how traditional QA and abuse-case thinking translate into effective AI security practices.

I document both what works and what fails, because realistic security depends on understanding both.

Recent work focuses on detecting and mitigating crescendo-style multi-turn attacks, including cross-turn risk accumulation, decay, and post-mitigation adversarial replay.

I focus on why AI systems fail over time, not just how they fail on a single prompt. My work shows how multi-turn interaction, context accumulation, and misplaced trust lead to real-world security breakdowns that single-turn testing routinely misses.

I design and execute structured AI red-team exercises against LLM-powered systems, emphasizing realism, documentation, and defense-in-depth rather than one-off prompt tricks.

My work focuses on:

• Prompt-based adversarial testing (memory evasion, authority abuse, policy contradiction)
• Multi-turn adaptive attack scenarios, including crescendo-style escalation and cross-turn pressure testing
• Canonical denial design and statelessness enforcement
• Training data claim suppression
• Evidence-based mitigation and residual risk analysis
• Stateless semantic risk scoring and escalation thresholds
• Cross-turn risk accumulation with deterministic decay (no conversational memory)

A full end-to-end red-team exercise designed to expose multi‑turn and orchestration failures in a locally hosted LLM API…, following a staged methodology similar to internal security assessments.

What this project demonstrates:

• Threat modeling for LLM-backed services
• Adaptive adversarial prompt design
• Multi-turn pressure testing
• Code-level policy enforcement
• Post-mitigation retesting
• Formal security documentation

Key Artifacts Produced:

• Stage-based findings reports
• Mitigation and risk assessment plans
• Residual risk register
• Security posture and controls matrix
• Known limitations and non-goals
• Adversarial evasion analysis and “where I would attack next” assessment

🔗 Repository: https://github.com/rodneystanley2025/ai-red-team-lab

Scans infrastructure and application configuration files (YAML, JSON, INI, TOML, etc.) for OWASP Top 10–style misconfigurations using a combination of rule-based analysis and AI-assisted review.

• Identifies insecure defaults, excessive permissions, exposed secrets, weak encryption, and dangerous functions
• Combines deterministic rules with LLM-assisted analysis
• Built with Python, Pydantic, LangChain, and open-source security rulesets
• Designed for extensibility and future CI/CD integration

🔗 Repository: https://github.com/rodneystanley2025/SecureConfigAI

• AI Red Teaming Labs – AI Red Teaming Labs – Prompt injection, memory evasion, authority abuse, crescendo escalation, and policy contradiction testing
• API Security Testing Lab – Reconnaissance and testing with Burp Suite, Postman, Nmap (APISec University)
• Vulnerability & Abuse-Case Testing – Volunteer Security QA at Chance AI (Dec 2024 – Mar 2025)
• Python & Security Automation – 400+ consecutive days on CodeSignal Learn (top percentile rankings)

• CompTIA Security+ (SY0-701) – 2025
• Cisco CCST Cybersecurity – 2025
• Certified Scrum Master (CSM) – 2024
• ISTQB Certified Tester Foundation Level (CTFL) – Since 2007
• Certified AI Security Expert (Msec-CAIS) – In Progress

AI & Security:
LLM Security, AI Red Teaming, Prompt Injection, Crescendo Attacks, Abuse-Case Testing, Defense-in-Depth

Security Tools:
Burp Suite, Nmap, Wireshark, Postman

Cloud & Infrastructure:
Azure, AWS, Docker, Kubernetes

Testing & Automation:
Selenium, C#, Azure DevOps, JIRA

Python & AI Tooling:
Python, Pydantic, LangChain, OpenAI/Groq APIs, Pandas

I’m always open to conversations about:

• AI red-teaming and LLM security
• Secure SDLC and abuse-case testing
• Career transitions into cybersecurity and AI security

🔗 LinkedIn
📧 rodney_stanley@hotmail.com

⭐️ From Rodney Stanley