Architecture and design
Clean architecture, design patterns, and SOLID. I've moved a monolith to microservices and designed serverless services on AWS.
15+ years of engineering fundamentals are what make AI output production-ready. I use AI agents to move faster, and I set up the codebase so they stay reliable. A person still reviews every change.
I know what good looks like, so I can guide agents and catch what they get wrong.
Clean architecture, design patterns, and SOLID. I've moved a monolith to microservices and designed serverless services on AWS.
Least-privilege access, OAuth 2.1, JWT, and audit trails. I've brought a Secure Development Lifecycle into a team and fixed access-control flaws in a shipped product.
Unit and integration testing, feature flags, testing with real users, analytics, and audit trails. Shipping is the middle, not the end.
Data structures, algorithms, and debugging down to the right layer, from C/C++ on retail hardware to APIs and the browser.
I set up the codebase so agents get it right more often, plus one rule that never changes.
Conventions, architecture notes, and project instructions written down where agents look for them. Agents stop guessing how things are done here, and their code fits the codebase.
Claude Code skills that capture how the team does recurring tasks, step by step. Every agent run follows the same proven path instead of improvising.
Tickets with clear scope, affected files, and acceptance criteria. When the task is well defined, the first draft is far closer to done.
MCP servers that give agents safe access to real tools and data, read-only by default. Checks agents can run on their own work before a person ever looks at it.
Agents draft, I review and test, and I decide what ships. A good setup means fewer corrections, not fewer reviews.
Agent-ready work in practice. A Claude Code plugin I built and open-sourced. It scores a Linear ticket for agent-readiness, fills the gaps from the codebase, and writes a spec. A person checks it before anything goes back to Linear.
Read-only before write access. Feature flags and real-user testing before launch. Then expand.
Build on the auth, tools, and components that already exist. Less new code, less risk.
I fix bugs where they start. I once proposed a new API field instead of having the frontend guess.
AI drafts and suggests. A person reviews, tests, and approves. AI-written code gets the same bar as mine.
Team feedback, analytics, and funnel tracking show what to build first and whether it worked.
Fully remote and async, so I write things down, and I turn workflows that work into tools the whole team can use.
Questions about my work or building with AI? Happy to chat.