I came up through physical systems, networks, and cloud control planes. That path made me practical: map the real system, find the risky edges, reduce the blast radius, and leave behind something a team can operate.
I work on cloud security, agent tooling, and the operational edges where clean designs become hard to reason about: permissions, state, logs, evidence, rollback, and human handoff. The AI work matters because teams are wiring language models into workflows faster than they are defining boundaries.
The thread has stayed the same: systems fail at boundaries. Cables, routes, IAM policies, logs, queues, humans, agents — the names change, but the work is still making the hidden parts visible enough to reason about.
I have spent enough time close to cloud control planes, incidents, migrations, and customer-facing operations to distrust clean diagrams. The interesting parts are usually permissions, state, quotas, network paths, logs, handoff, and rollback.
Before changing a system, map what exists, where it fails, and what evidence proves the fix.
Good engineering is not optimism. It is permissions, scopes, defaults, tests, and rollback paths.
The output should be something a team can use: findings, code, runbooks, decisions, or a working integration.
Cloud security, migration guardrails, network-heavy architecture, security automation, MCP integrations, compliance-minded workflows, and founder-level technical execution are good fits when the goal is a useful artifact rather than theater.