Portfolio
Whitepapers, novel research, and selected engagements.
Research, open-source contributions, and frameworks from the Brasstower bench — the advisors and specialists who deliver client engagements. This work spans AI and agentic security, cryptographic protocol design, privacy-preserving ML architectures, and security automation at scale.
Brasstower’s bench has helped build security at:
50+
Granted Patents
125+
Combined Years in Security
5
Industry Standards Contributions
Research & Publications
Verifiable Semantics for Agent-to-Agent Communication
A certification protocol that bounds semantic disagreement between agents via core-guarded reasoning. Addresses how AI systems achieve true meaning across communication boundaries.
Read on arXiv →
... and on LinkedIn →
Securing Multi-Tool AI Agent Chains with Dynamic, Real-Time Compositional Policies
Introduces the Dynamic Security Context Composition system and Monotonic Resolution Satisfiability algorithm for formal policy composition across multi-tool agentic workflows.
arXiv link pending →
Google Secure AI Framework (SAIF)
Industry framework for securing AI systems across the development lifecycle. Developed during tenure as Head of Security Engineering and Platforms at Google DeepMind.
Separable Expert Architecture (SEA)
Per-user LoRA adapters enabling GDPR-compliant model personalization. Users can delete their data at any time and model behavior updates without exposing user data or retraining shared weights.
Frameworks & Architectures
SAEF — Secure Agentic Execution Framework
Zero-trust execution framework for AI agents. Formal policy composition, stigmergic data store, and three-plane audit separation. Built on verifiable language semantics.
Green Code
Semantic cache for reviewed code blocks. Reduces redundant security review overhead by identifying previously-reviewed code patterns across repositories.
Open Source & Industry
android-security-lints
Automated security lint checks for Android applications. Part of a broader automated remediation program that scaled security fixes from ~300/year to ~30,000/month.
View on GitHub →
Android Developers Blog
Published guidance on Android security development practices and automated security tooling for the Android developer ecosystem.
Standards Contributions
ISO 27090
AI Security · Contributor
AI Security · Contributor
ISO 27091
AI Privacy · Contributor
AI Privacy · Contributor
OWASP ML Top 10
Machine Learning Security
Machine Learning Security
Google SAIF
Secure AI Framework · Co-author
Secure AI Framework · Co-author