AI Engineer
Building and Evaluating AI Architecture | Helping teams to leverage AI
Day to day I work on RAG architecture and evaluation, MCP server orchestration, and internal AI tooling. A good chunk of my time also goes to enablement — I'm a Claude Code Enablement lead and host training across the org. I also consult with teams on their skills, agent loops, and workflows and help them enhance and scale.
Most days fall into three buckets:
- Guardrails & infra — eval harnesses, merge-gate pipelines, and feedback loops that catch hallucinations in-loop and correct them before they ship. Cuts cycle time and saves tokens.
- Agentic architecture — I sit with internal teams, learn their workflows, figure out what's actually worth automating, then build the agent skills, mcps, and architecture to do it within their constraints (cost and scale included).
- Mentorship — I help teams of all technical backgrounds get fluent with AI tools, and I interview and train junior AI Engineers and SDETs on engineering fundamentals. This is the part I like most.
Most of the projects on here are me proving out an idea, building a lightweight POC to get buy-in for something at work, or just scratching my own itch. I care more about the patterns and lessons than making things pretty for the most part, though occasionally I will scale something if I find value.
A couple examples of this are:
Agentic Playwright Team, a Claude Code-native test-generation harness for Cucumber.js and Playwright. I built the original as a POC in one night to get buy-in from QA teams at work, then ended up bringing it in and scaling it to a full agentic harness.
Playback MCP, an MCP server + Chrome Extension pair that lets MCP clients manipulate and process YouTube videos. This one started as a bespoke monolith I built for driving my guitar practice sessions, but a lot of my network was really interested in the orchestration piece so I refactored and published the tooling. It's live on the Chrome Store and NPM!
I occasionally write a blog post on my HashNode blog, usually on agent tooling, context engineering, and whatever architecture or ideas I'm playing around with.
I also publish a YouTube video here and there when I want to hear myself talk. One cool series is my Claude Code from Zero playlist, a YouTube course that ended up seeding a lot of the enterprise training materials I help with.
Reach out here or on any of my socials — always happy to talk shop, trade ideas, or just connect with people building cool things.



