Trust deliberately.
Give the right access, with clear boundaries, and assume nothing by default.
AI SECURITY · INFRASTRUCTURE · SYSTEMS CLARITY
Security architecture
you can actually operate.
I help teams use AI and modern tooling safely — with clear tool-calling boundaries, human approvals, comprehensive logging, and security systems designed for the real world.
01 / Our philosophy
A clear standard for every engagement.
Enforce predictably.
Use consistent controls, approvals and monitoring so systems behave as expected.
Keep systems understandable.
Design for clarity, with simple patterns, good documentation, and observable behaviour.
// Services
Practical security for modern systems
AI Security & Governance
Define safe tool-calling boundaries, approval flows, and auditability for AI systems.
Systems Architecture
Design and review secure, maintainable infrastructure and application systems.
Security Reviews
Assessments, threat modelling, and practical recommendations that reduce risk.
Implementation Support
Turn plans into working systems with clear documentation and knowledge transfer.
// Selected engineering work
Methods you can inspect.
Independent projects that make authority, policy changes, and decision evidence easier to examine. Each walkthrough uses a synthetic example with reproducible inputs and results.
Inspect declared authority
Coldgate
See weakened approval requirements and wider declared access in a saved configuration comparison.
Read the Coldgate walkthrough →Test policy changes
Driftglass
Find newly allowed actions even when the recorded workflow still passes.
Read the Driftglass walkthrough →Inspect the evidence
SignalForge
Separate repeated coverage from independent support and keep counter-evidence visible.
Read the SignalForge walkthrough →// Principles
Simple ideas. Durable systems.
Least privilege, everywhere.
Constrain access, scope tools, and reduce blast radius.
Observable and auditable.
Log what matters. Make system behaviour understandable.
Human judgment in the loop.
Automate confidently, with clear approval points for meaningful actions.
// Reference architecture
A safe tool-calling pattern
User
or Agent
or Agent
Policy &
Approval Layer
Approval Layer
Tool Gateway
(Allow / Deny)
(Allow / Deny)
Approved ToolsInternal SystemsExternal Services
▧ Log all requests, responses and decisions
// Engagements
Focused scope. Practical outcomes.
AI Tool-Calling Guardrails
Design / Implementation / Documentation
Approval flows and logging for LLM tool use across internal systems.
Infrastructure Security Review
Assessment / Threat modelling / Roadmap
Cloud infrastructure review, prioritised fixes and a phased improvement plan.
Internal Developer Platform
Architecture / Implementation / Enablement
Secure, self-service internal tools with clear access controls.
// Contact
Let’s build systems you can trust.
Have a security or systems
challenge in mind?
I work with organizations in Victoria, BC and remotely to design, review, and implement practical, secure systems.
FROSTEDLOGICSecurity and systems practice
Victoria, BC