How My Portfolio Site Became an AI Project

The Domain That Sat There For Years

I registered the domain years ago. For 15 years, steady employment meant I never needed a personal site. The resume lived in a Google Doc. The portfolio lived in my head. There was always a good reason to wait.

The interest in AI and machine learning had been building for a while, though. I completed an AI/ML bootcamp while still employed, earning a certificate and finishing a capstone project on Google Cloud Platform. The itch was real, but the day job kept the pace slow.

Then the layoff hit. Without warning and without connection to performance. I had taken care of my team, shipped on time, managed the complexity. None of it mattered when the decision came down. That is the part nobody prepares you for.

The first few days after felt strange. I had a fresh AI/ML certificate, no job, and a domain pointing at nothing. The decision was obvious.

Starting With What I Know

I built the site the way I would build any engineering project at work. Next.js 14 with the App Router, TypeScript in strict mode, Contentful as the headless CMS, Tailwind for styling, and Netlify for hosting and CI/CD. A tech stack I trusted. An architecture I could reason about.

The initial goal was minimal. A resume page, a portfolio section, and maybe a blog. Clean code, separation of concerns, tested services. I wrote a CLAUDE.md that reads like an engineering standards document, covering architecture principles, rendering strategies, and what counts as done. Not because anyone required it, but because that is how I think about systems.

The habit of treating personal projects like production systems is something I picked up after years of watching shortcuts compound into rewrites. Good architecture at the start costs nothing compared to the cleanup later. That lesson transfers.

When AI Crept In

The first shift happened when I started using Claude Code to accelerate development. That went well enough that I kept going. I built CLI commands for job matching, cover letter generation, and company research scanning. Then I built skills and agents that enforce writing style and scaffold new components with the right patterns baked in.

My capstone ML project from the bootcamp, Tactical Portfolio Intelligence (TPI), became a portfolio piece on the site. It is a classification system built on GCP, and getting it out of a notebook and into a published case study forced me to articulate the actual engineering involved. That process was valuable.

The next step was a RAG-powered chatbot. I connected OpenAI embeddings to a Pinecone vector store, added multi-provider LLM support, and built a retrieval layer that can answer questions about my experience and projects. The system pulls from real documents, not hallucinated summaries. That is a meaningful distinction.

At some point I looked up and realized more than half the interesting engineering on the site was AI work. That was not the plan.

The Accidental Pivot That Was Not Really Accidental

Looking back, the pivot had been in motion before the layoff. The bootcamp was step one. The layoff just removed the last reason to move slowly.

Each problem I solved on the site led to a more AI-native solution. Job matching became semantic scoring with embedding similarity rather than keyword overlap. The resume system became a content model with per-profile bullet selection, where different roles see different highlights from the same underlying experience. The chatbot became a real retrieval pipeline with proper chunking and context management.

The site itself is now the strongest AI/ML signal on my resume. Not because I listed the skills in a sidebar, but because I shipped real systems. A semantic job matcher, a vector-backed chatbot, a cloud ML pipeline. Those are concrete artifacts. A recruiter or hiring manager can read the case studies and see exactly what I built and how.

The lesson I keep coming back to: the best career pivots happen when you have been quietly preparing, then life gives you the push. The push was abrupt and unwelcome. The preparation made it matter.

Build the Site

If you have owned a domain for years and never built anything on it, build something now. Do not wait for the right moment or the complete plan. Neither will arrive on schedule.

Start with what you know. Solve the real problems in front of you. The work will show you where to go next.