Specify the contract
Make inputs, invariants, ownership and failure behavior explicit before automating changes.
Independent systems engineer
Kubernetes & AI · Security-conscious automation · Open source
I build code and infrastructure at the boundaries of Kubernetes, AI, networking and security. Explore working implementations, explicit failure paths and documented measurements—not just architecture diagrams.

Make inputs, invariants, ownership and failure behavior explicit before automating changes.
Reject stale evidence, unsupported changes and invalid output; exercise recovery as well as success.
Link code, tests and dated measurements. Label reference designs and roadmap work as such.
Open source / curated implementation evidence
Each project identifies what is implemented, where to inspect the evidence and what remains unverified. The complete public repository stream is available in the project index.
A synthetic Kubernetes incident replay connecting an unexpected shell to a blocked outbound connection, with inspectable evidence and human review boundaries.
A subscription-free delivery reference that tests Azure Terraform mock plans with OPA, scans source and runtime images, and makes every mandatory CI gate explicit.
An offline EC2 quota incident demonstrator with deterministic diagnosis, policy-bounded Terraform proposals, SQLite audit history and recovery verification.
A FastAPI and Redis queue that wakes CPU workers and GPU vLLM pods with KEDA, with interrupted-job replay and a GKE deployment guide.
Provider-neutral zone validation, record diffs, snapshots, recovery attempts, propagation checks and read-only IPAM lookup across Cloudflare, Route 53, Google Cloud DNS, Azure DNS and TCPWave.
A Terminal-Bench log-report task hardened with a digest-pinned container, removal of a leaked reference solution, and an independent verifier that recomputes metrics from the source log.
A Helm-based deployment of an Elixir Phoenix application with PostgreSQL, environment configuration, database migrations, health probes and public ingress.
A local-first workflow policy scanner and cost-report CLI with guarded provider examples, multi-cloud Terraform, Helm deployment and observability starters.
Technical expertise
Automation / systems in practice
Real examples from the repositories and field notes, described without invented hours-saved or performance claims.
Validate records, compare desired and live state, snapshot before writes, and check propagation after the change.
DNS migration automationHow I connected Terraform mock plans, tested OPA guardrails and runtime image scans without requiring a cloud subscription or claiming a deployment.
Read the noteA quota incident demo that separates evidence, diagnosis, bounded proposals and verified recovery—with a manual path when AI is unavailable.
Read the noteExperience / systems in practice
Project-backed engineering practice—not an invented employment history. Follow the code and design notes for scope, tradeoffs and failure modes.
Capacity, scheduling and rollout practices for systems that move models from experiment to production.
DNS and IPAM workflows that make changes reviewable, validated and recoverable.
Observable deployment paths with a clear control, a decision rule and a rollback plan.
Typed inputs, refusal paths, bounded proposals and recovery checks—with source code and explicit limitations.
Tools used in code and reference designs
Contact / let's collaborate
Open to conversations about AI infrastructure, platform engineering, networking and safer operations.