contact
Start with the boundary, the risk, and the system goal.
Useful work starts when the context is concrete: what is being shipped, what can break, what data or tools are involved, who owns the service, and what proof would make the decision easier.
best fit
Where I can help
- Secure AI feature design: model traffic, retrieval, tools, policy gates, evals, and trace review.
- Platform delivery cleanup: Kubernetes readiness, CI/CD gates, release notes, rollback, and observability.
- Application and API security review with evidence that separates real impact from noise.
- Architecture writing for teams that need crisp boundaries, tradeoffs, controls, and handoff.
- DevSecOps operating models that connect engineering velocity with practical risk reduction.
working boundary
Make the first message useful
- No secrets or customer data should be sent by email.
- Security testing needs explicit authorization, scope, and expected boundaries.
- Private reports stay private; portfolio artifacts stay sanitized and representative.
- Helpful context includes stack, owner, deadline, data sensitivity, current pain, and what decision is blocked.
signal
What to include
- AI / LLM
Model provider, retrieval source, tool permissions, risky outputs, eval coverage, and logging gaps.
- Security
Authorized scope, affected role or tenant boundary, reproduction constraints, and desired report depth.
- Platform
Runtime, deployment path, release pain, rollback expectations, observability gaps, and service ownership.