Projects

Current focus

• SenTeGuard: A security layer for enterprise GenAI that detects, warns, flags and blocks risky content flows at the point of use, especially semantic leakage when employees summarize, paraphrase, translate, or abstract sensitive material.
• Moyo: An "information-space mapper" that crawls and maps an organization's public-facing knowledge surface, then measures how close internal concepts are to becoming externally reachable through LLMs, search, and retrieval tools. Used for outside-in risk discovery and targeted AI red-teaming that feeds directly into concrete controls and SenTeGuard deployments.
• Joseki:Wrapperhub: A prompt-spec marketplace and version-control layer for reusable AI workflows. The goal is to make prompts and agentic "recipes" auditable, testable, and governable: with changelogs, licensing, security scoring, and organization-level guardrails for how prompts are shared and reused.
• Idea Security Verification Framework (ISVF): An open framework for defining, testing, and auditing inference boundaries in LLM-enabled systems — covering semantic leakage, cross-domain inference, reachability risks, and evidence-pack requirements for enterprise compliance.