Building this portfolio and résumé with LLMs
I used Claude Code and Codex to build this portfolio and its résumé. I supplied the facts, explained what I wanted, corrected drafts and checked the finished pages and PDFs before publishing.
Results
2résumé formats from the same verified history
3site themes checked against shared content
The problem
My work history and technical evidence were spread across notes and projects. I needed to turn that material into short, accurate explanations for employers, with checks to keep private information out of the site.
What I did
- Used linked Obsidian notes, project decisions and session handoffs to give the models the history and constraints for each task.
- Used Superpowers brainstorming to work out requirements before implementation. Reusable agent instructions carry the rules between sessions. I correct drafts that miss the tone, scope or facts.
- Used adversarial review by asking Claude and GPT to challenge plans and drafts against shared rubrics. I investigate disagreements and check the evidence before accepting a change.
- Built this portfolio, a designed résumé and an ATS version with AI coding tools, then checked confidentiality, builds, browser behavior and PDF layouts.
- Used methods tested at work, including headless agents with approval checks and deny hooks, a create-only vault gardener, and AI-assisted Blender iterations with recorded reviews and failures.
What I took from it
The models need the relevant notes and a clear description of what I'm trying to do. I correct them when they miss something, test their changes and take responsibility for what I publish.