Liam Romanis

Offensive security specialist: red teaming, pentesting, AD/Azure/AWS/M365/K8s. Python, PowerShell, Bash. AI/ML security, prompt injection, LLM supply chain. 10+ years practitioner background.

Advancing AI/ML security tooling — LLM dataset sanitisation, prompt injection detection, and AI supply chain security. Building with the Anthropic API and exploring AI agent attack surfaces.

DataGuard — an LLM training dataset security scanner. Also building AI-powered security tooling: HuggingFace model scanners, forensic agents, and prompt injection detection frameworks.

Collaborations on offensive security tools, AI/ML security research, and dataset integrity work. Always happy to connect with AI red teamers, security researchers, or anyone building AI tooling.